PERINATAL MORTALITY-System Related and Environmental Factors-

Virender Jashvant John Poeran

Perinatal Mortality

– System Related and Environmental Factors –

Virender Jashvant John Poeran

Perinatal Mortality – System Related and Environmental Factors –

PhD thesis, Erasmus University Rotterdam, The Netherlands

Cover design Joseph Mallchok | Maaike van der Staal, YourThesis

Layout Renate Siebes, Proefschrift.nu

Printing Ridderprint, Ridderkerk

Publisher Medix Publishers BV, Keizersgracht 317A, 1016 EE Amsterdam

ISBN/EAN 978-94-91487-08-8

Acknowledgements

The board of The Netherlands Perinatal Registry kindly provided permission to use the

Registry for research purposes.

The research presented in this dissertation was performed at the department of Obstetrics

and Gynaecology, division of Obstetrics and Prenatal Medicine, Erasmus MC, Rotterdam,

The Netherlands

Financial support for this dissertation was kindly provided by:

The department of Obstetrics and Gynaecology Erasmus MC, The Rotterdam municipal

authorities and the ‘Ready for a Baby’ programme (www.klaarvooreenkind.nl), J.E. Jurriaanse

Stichting, Medical Dynamics, Mpluz, Danone, BMA-Mosos, Goodlife Pharma, AbbVie BV,

Nutricia, Vasa Previa Foundation (www.vasaprevia.nl).

“The Dutch Vasa Previa Foundation is a non-profit organization whose primary objective is

the reduction of infant mortality or permanent injury due to vasa previa. Some activities to

achieve this are drawing attention to the risk of vasa previa within the relevant professional

groups and aiming for adjustment of medical protocols, which are important for diagnosis and

management of vasa previa. In addition, the foundation offers support and advice to anyone

who is experiencing or has experienced vasa previa.”

Copyright © 2013

Virender Jashvant John Poeran, Rotterdam, The Netherlands, v.j.j.poeran@gmail.com

All rights reserved. No part of this thesis may be reproduced in any form or by any means

without written permission from the author.

Perinatal Mortality

– System Related and Environmental Factors –

Perinatale sterfte: systeem gerelateerde

en omgevingsgebonden factoren

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de

Erasmus Universiteit Rotterdam

op gezag van de rector magnificus

Prof.dr. H.G. Schmidt

en volgens besluit van het College voor Promoties.

De openbare verdediging zal plaatsvinden op

woensdag 18 september 2013 om 11:30 uur

door

Virender Jashvant John Poeran

geboren te Paramaribo (Suriname)

PROMOTIECOMMISSIE

Promotoren Prof.dr. G.J. Bonsel

Prof.dr. E.A.P. Steegers

Overige leden Prof.dr.ir. A. Burdorf

Prof.dr. A.J. van der Heijden

Prof.dr. A.P. Verhoeff

Copromotoren Dr. E. Birnie

Dr. S. Denktaş

Paranimfen Drs. Jacoba van der Kooy

Drs. Martijn Wagemans

CONTENTSChapter 1 Introduction 7

Part I: Perinatal health in Rotterdam

Chapter 2 Urban perinatal health inequalities 17

Chapter 3 Pregnancy and birth in Rotterdam: a local health report 25

Chapter 4 Social deprivation and adverse perinatal outcomes among Western and non-Western pregnant women in a Dutch urban population

41

Part II: Perinatal health in The Netherlands

Chapter 5 Population attributable risks of patient, child and organisational risk factors for perinatal mortality

61

Chapter 6 The ‘Healthy Pregnancy 4 All’ Study: design and cohort profile 83

Chapter 7 The effectiveness of risk selection in the Dutch obstetric care system

111

Chapter 8 Planned home compared with planned hospital births in The Netherlands: intrapartum and early neonatal death in low-risk pregnancies

129

Chapter 9 Does centralisation of acute obstetric care reduce perinatal mortality? An empirical study of over 1 million births in The Netherlands

149

Chapter 10 The impact of extremes in outdoor temperature and sunshine exposure on birthweight

171

Chapter 11 General discussion 191

Chapter 12 English and Dutch summary 207

Chapter 13 Authors and affiliations 215

Manuscripts 217

About the author 220

PhD portfolio 222

Word of thanks 224

1Introduction

8

INTRO

DU

CTION

1 In The Netherlands perinatal mortality rates exceed the European average.1-3 On a second

geographic level of comparison, i.e., within The Netherlands, adverse perinatal outcome

rates are much higher in the four largest cities (‘G4’, i.e., Amsterdam, Rotterdam, The Hague,

Utrecht).4,5 Again, on a third level, i.e., within the G4-cities, adverse perinatal outcomes are

overrepresented in socially deprived areas on the borough- and neighbourhood level.6-9

For long, population factors such as the high age of mothers at first childbirth, the high

prevalence of multiple pregnancies (as a consequence of either assisted reproduction or

high maternal age), and the increasing prevalence of non-Western pregnant women were

held responsible for the high perinatal mortality.1,2 However, these explanations were

challenged as perinatal mortality remains high in analyses after exclusion of these risk

groups.10 Recent studies have thus addressed the potential role of other factors, in particu-

lar healthcare related factors and geographic (e.g., neighbourhood, environment) factors.

Healthcare related factors put forward the unique system of Dutch obstetric care with

independently practicing community midwives11,12, travel time to hospital13, and organi-

sational characteristics of hospitals14. Candidate environmental factors are physical factors

(e.g., air pollution15,16 and ambient noise pollution17), and aggregate social factors like urban

deprivation4,7,8,18.

As stated, three geographic levels of inequalities in perinatal health are present, i.e., (1) The

Netherlands vs. Europe, (2) regions and G4-cities vs. the remainder of the Netherlands, and

(3) deprived vs. non-deprived neighbourhoods and boroughs within these G4-cities. The

causal factors of perinatal inequalities may differ between the three geographic levels. An

important cause (e.g., social deprivation) at one level may play an insignificant role at another

level. Furthermore, as the majority of studies only deliver relative effect measures5,6,8,11-14,19

with or without adjustment such as odds ratios, there is insufficient information on the

importance of risk factors for perinatal mortality in absolute terms on the population level

(e.g., population attributable risks), and the effect size of interdependence (interaction)

between these risk factors. An estimate of attribution, including interaction, is critical to

prioritise among options available to tackle perinatal mortality.

Finally, the definition of perinatal mortality has to be taken into account: death from 22

weeks of gestational age until 7 days postpartum. This definition combines stillbirth and

neonatal mortality; hence, perinatal mortality differences may rest on different underlying

patterns of these two components. Organisational studies within the Netherlands focus on

intrapartum and neonatal mortality14, but international comparison shows that in particular

(term) stillbirth mainly contributes to the poor Dutch position on perinatal health.1-3

9

INTRO

DU

CTION

1In response to the high perinatal mortality, the Dutch Minister of Health issued several

measures20 in 2008, of which two were deemed most important: (1) the appointment of an

advisory board on pregnancy and birth, and (2) the request for a national report on priority

setting of perinatal research. The advisory board proposed several recommendations in their

final report (‘A good beginning’) in 2010.20,21 Two important recommendations refer to (1)

targeting pregnancy and deprivation (particularly in urban areas), and (2) universal access

within 15 minutes to ‘qualified professionals’ (midwives, gynaecologists, paediatricians,

anaesthesiologists, and operating theatre staff ) to guarantee the quality of 24*7 acute

obstetric care.

Most recommendations from the advisory group report were also mentioned in the research

topics proposed in the ZonMw (The Netherlands Organisation for Health Research and

Development) Signalement study, which provided the quantitative background to some of

the aforementioned recommendations.22 In this study, the unique Dutch system of obstetric

care was addressed on patient, environment and healthcare related risk factors of perinatal

mortality and morbidity. Also, the concept of ‘Big4’ was introduced; ‘Big4’ refers to four

adverse pregnancy outcomes (perinatal morbidities) which precede perinatal mortality

in 85% of cases.22 ‘Big4’ morbidities were defined as the presence (single or combined)

of congenital anomalies (list defined), preterm birth (<37th week of gestation), small

for gestational age (SGA, birthweight below the 10th percentile for gestational age23) or

low Apgar score (<7, 5 minutes after birth). Mostly, the first three ‘Big4’ morbidities offer

some interventional opportunities in terms of diagnosis, preventive actions or treatments

when timely detected. Figure 1.1 illustrates in a Venn diagram the relationship between

(combinations of ) ‘Big4’ morbidities and perinatal mortality. Mainly, the presence of more

than one ‘Big4’ morbidity poses a significant risk for perinatal mortality, e.g., a low Apgar

score combined with preterm birth occurs in 30.3% of all cases of perinatal mortality.22

The Signalement study proposed a ‘stepwise’ model to represent the causal chain to

perinatal mortality.22,24 The occurrence of ‘Big4’ morbidities (‘step 1’) and their progression

to perinatal mortality (‘step 2’) is influenced by the accumulation18 and interaction of

patient, environment and healthcare related risk factors. This ‘stepwise’ model not only offers

different windows for intervention, but also implies different ways to intervene in each step.

In addition to national policy measures to increase perinatal health, local and regional

projects were launched. One comprehensive regional project is the ‘Ready for a Baby’

programme in the multi-ethnic city of Rotterdam.25,26 This public health programme was

instituted by the municipal council of Rotterdam, with collaboration of the local public health

10

INTRO

DU

CTION

1

authorities (GGD Rotterdam Rijnmond), the university medical centre (Erasmus MC) and

the healthcare professionals (e.g., gynaecologists, midwives, youth healthcare physicians).

The main object of this 10-year programme is to improve perinatal health and to reduce

perinatal mortality in all districts of Rotterdam to at least the current national average.25,26

Key elements include: (1) better understanding of the large health differences between

women living in deprived and non-deprived urban areas5,6, (2) improving cooperation

between healthcare professionals in the chain of obstetric healthcare27, (3) developing

improved methods for risk selection before and during pregnancy, and (4) methods to reach

low educated and immigrant groups. Baseline measurement and continuous monitoring

of the effect of interventions in the ‘Ready for a Baby’ programme are vital in this process.

Figure 1.1 Venn diagram with the prevalence per 1,000 births (¶) of separate and combined ‘Big4’ morbidities and their contribution to all cases of perinatal mortality († in %); this adds up to 85% of all cases of perinatal mortality.

¶ 0.3† 1.9%

CONGENITALANOMALY

PRETERMBIRTH

SMALL FORGESTATIONAL AGE

LOW APGARSCORE

¶ 0.7† 6.7%

¶ 3.1† 0.9%

¶ 54.8† 1.5%

¶ 0.6† 2.3%

¶ 62.9† 0.2%

¶ 3.2† 0.2%

¶ 18.0† 1.2%

¶ 6.4† 3.2%

¶ 1.5† 5.6%

¶ 1.0† 7.0%

¶ 2.5† 23.4%

¶ 0.4† 0.2%

¶ 5.1† 30.3%

¶ 2.2† 0.4%

11

INTRO

DU

CTION

1AIM OF THIS THESISThe aim of this thesis is to investigate the main contributing factors in adverse perinatal

outcomes on the three geographic levels. In particular the relative role of patient-,

environment- and healthcare related risk factors is studied in order to select rational health

policies and to decrease gaps in perinatal health.

This thesis has two parts, each of which draws evidence from different sources. Part I

describes results from the regional ‘Ready for a Baby’ programme, and part II contains

studies on the national level, either ordered by ‘ZonMw’ (The Netherlands Organisation

for Health Research and Development) concerning the ‘Signalement’ study, ‘SAZ’ hospitals

(a co-operative institution of circa 40 small general hospitals in The Netherlands), or the

Ministry of Health, Welfare and Sport (regarding the national programme ‘Healthy Pregnancy

4 All’). These studies will guide the reader through the eight research questions, the first

two questions being addressed in part I (Rotterdam), and questions three to eight being

addressed in part II (The Netherlands).

Part I. Rotterdam

1. How are perinatal mortality and morbidity distributed within the city of Rotterdam?

2. To what extent are determinants of social deprivation pertinent to differences in perinatal

mortality and morbidity?

Part II. The Netherlands

3. What is the contributing role of patient (maternal and child) and non-patient (healthcare

organisation, environment) risk factors in perinatal mortality on the population level?

4. What is a rational method to select priority areas for improved prevention of adverse

perinatal outcome?

5. Is the Dutch system of risk selection adequate in separating low risk pregnancies from

high risk pregnancies?

6. Can the unique feature of home birth in Dutch obstetric care still be offered as a safe

option to a selected group of women?

7. Is centralisation of acute obstetric care an effective intervention, on the healthcare

organisational level, to improve perinatal outcome?

8. Do climatological factors provide an additional explanation for inequalities in perinatal

outcome?

12

INTRO

DU

CTION

11. Buitendijk SE, Nijhuis JG. High perinatal mortal-

ity in the Netherlands compared to the rest of Europe. Ned Tijdschr Geneeskd. 2004;148:1855-1860.

2. Mohangoo AD, Buitendijk SE, Hukkelhoven CW, et al. Higher perinatal mortality in The Netherlands than in other European countries: the Peristat-II study. Ned Tijdschr Geneeskd. 2008;152:2718-2727.

3. The EURO-PERISTAT project in collaboration with SCPE EE. European perinatal health report, 2008. (online: www.europeristat.com).

4. Tromp M, Eskes M, Reitsma JB, et al. Regional perinatal mortality differences in the Netherlands; care is the question. BMC Public Health. 2009;9: 102.

5. de Graaf JP, Ravelli AC, Wildschut HI, et al. Perinatal outcomes in the four largest cities and in deprived neighbourhoods in The Netherlands. Ned Tijdschr Geneeskd. 2008;152(50):2734-2740.

6. de Graaf JP, Ravelli AC, de Haan MA, Steegers EA, Bonsel GJ. Living in deprived urban districts increases perinatal health inequalities. J Matern Fetal Neonatal Med. 2012 (in press).

7. Poeran J, Denktas S, Birnie E, Bonsel GJ, Steegers EA. Urban perinatal health inequalities. J Matern Fetal Neonatal Med. 2011;24:643-646.

8. Ravelli AC, Steegers EA, Rijninks-van Driel GC, et al. Differences in perinatal mortality between districts of Amsterdam. Ned Tijdschr Geneeskd. 2011;155:A3130.

9. Poeran J, Denktas S, Birnie E, et al. Pregnancy and birth in Rotterdam: a local health report (in Dutch: Zwangerschap en geboorte in Rotterdam: een lokale gezondheidsrapportage). Tijdschrift voor gezondheidswetenschappen. 2012;90:496-503.

10. Anthony S, Jacobusse GW, van der Pal-de Bruin KM, Buitendijk S, Zeitlin J. Factors E-PWGoR. Do differences in maternal age, parity and multiple births explain variations in fetal and neonatal mortality rates in Europe?--Results from the EURO-PERISTAT project. Paediatr Perinat Epide-miol. 2009;23:292-300.

11. van der Kooy J, Poeran J, de Graaf JP, et al. Planned home compared with planned hospital births in the Netherlands: intrapartum and early neonatal death in low-risk pregnancies. Obstet Gynecol. 2011;118:1037-1046.

12. Evers AC, Brouwers HA, Hukkelhoven CW, et al. Perinatal mortality and severe morbidity in low and high risk term pregnancies in the Netherlands: prospective cohort study. BMJ. 2010; 341:c5639.

13. Ravelli AC, Jager KJ, de Groot MH, et al. Travel time from home to hospital and adverse perinatal outcomes in women at term in the Netherlands. BJOG. 2011;118:457-465.

14. de Graaf JP, Ravelli AC, Visser GH, et al. Increased adverse perinatal outcome of hospital delivery at night. BJOG. 2010;117:1098-1107.

15. de Medeiros AP, Gouveia N, Machado RP, et al. Traffic-related air pollution and perinatal mortal-ity: a case-control study. Environ Health Perspect. Jan 2009;117(1):127-132.

16. Wilhelm M, Ritz B. Residential proximity to traffic and adverse birth outcomes in Los Angeles county, California, 1994-1996. Environ Health Perspect. Feb 2003;111(2):207-216.

17. Hartikainen AL, Sorri M, Anttonen H, Tuimala R, Laara E. Effect of occupational noise on the course and outcome of pregnancy. Scand J Work Environ Health. Dec 1994;20(6):444-450.

18. Timmermans S, Bonsel GJ, Steegers-Theunissen RP, et al. Individual accumulation of heterogene-ous risks explains perinatal inequalities within deprived neighbourhoods. Eur J Epidemiol. 2011;26:165-180.

19. Ravelli AC, Tromp M, van Huis M, et al. Decreas-ing perinatal mortality in The Netherlands, 2000-2006: a record linkage study. J Epidemiol Community Health. 2009;63:761-765.

20. Sheldon T. Prompt access to expert care is among advice to reduce perinatal deaths in Netherlands. BMJ. 2010;340:c277.

21. Van der Velden J. Dutch report: Een goed begin. Veilige zorg rond zwangerschap en geboorte. Advies Stuurgroep Zwangerschap en Geboorte. The Hague: 2009.

REFERENCES

13

INTRO

DU

CTION

122. Bonsel GJ, Birnie E, Denktas S, Poeran J, Steegers

EAP. Dutch report: Lijnen in de Perinatale Sterfte, Signalementstudie Zwangerschap en Geboorte 2010. Rotterdam: Erasmus MC; 2010.

23. Kloosterman GJ. Intrauterine growth and in-trauterine growth curves. Maandschr Kinderge-neeskd. 1969;37:209-225.

24. Poeran J, Steegers EA, Bonsel GJ. Approach to perinatal mortality in the Netherlands: outcomes of a systematic expert study. Ned Tijdschr Geneeskd. 2012;155:A4499.

25. Denktas S, Bonsel GJ, Steegers EA. Perinatal health in Rotterdam, the Netherlands--experiences after 2 years of ‘Ready for a baby’. Ned Tijdschr Geneeskd. 2012;156:A4289.

26. Denktas S, Bonsel GJ, Van der Weg EJ, et al. An urban perinatal health programme of strategies to improve perinatal health. Matern Child Health J. 2011;16:1553-1558.

27. Posthumus AG, Scholmerich VL, Waelput AJ, et al. Bridging Between Professionals in Perinatal Care: Towards Shared Care in The Netherlands. Matern Child Health J. 2012 (in press).

14

INTRO

DU

CTION

1

PART I

Perinatal health in Rotterdam

16

2

Journal of Maternal-Fetal and Neonatal Medicine. 2011;24:643-646.

Urban perinatal health inequalities

J. Poeran, S. Denktaş, E. Birnie, G.J. Bonsel, E.A.P. Steegers

18

URBA

N PERIN

ATAL H

EALTH

INEQ

UA

LITIES

2

ABSTRACTObjective Large urban areas have higher perinatal mortality rates. In attaining a better

understanding, we conducted an analysis on a neighbourhood level in Rotterdam, the

second largest city of The Netherlands.

Methods Perinatal outcome of all single pregnancies (50,000) was analysed for the period of

2000-2006. The prevalences of perinatal mortality and perinatal morbidity were determined

for every neighbourhood.

Results Large perinatal health inequalities exist between neighbourhoods in the city of

Rotterdam with perinatal mortality rates as high as 37 per 1,000 births. The highest risks

were observed in deprived neighbourhoods.

Conclusion We observed high levels of perinatal health inequalities in the city of Rotterdam

which have not been previously described in the Western world. Accumulation of medical

risk factors as well as socio-economic and urban risk factors seems to be a likely contributor.

19

URBA

N PERIN

ATAL H

EALTH

INEQ

UA

LITIES

2

INTRODUCTIONHealth inequalities pose a complex problem in the field of public health, consisting of

medical, social and socio-economic aspects. Considering the substantial long term effects,

adverse perinatal health outcomes, along with perinatal health inequalities, are of particular

interest.1,2

In The Netherlands, perinatal mortality exceeds the European average despite a high

standard of mother and child healthcare.3,4 A recent national study showed an increased

risk of 12% on adverse perinatal outcomes in large urban areas; for non-Western women

this additional risk was 26%.3 The largest risk was observed in deprived neighbourhoods

where ethnicity was expected to be a key etiologic factor.3 However, the effect of living in

deprived neighbourhoods, as compared to living elsewhere, was most prominent among

Western women with an increased risk of 24%.3 In attaining a better understanding of this

complicated matter, we conducted an analysis on the neighbourhood level in the second

largest city of The Netherlands, i.e., the city of Rotterdam where socio-economic inequalities

are most obvious.

METHODSThe Netherlands Perinatal Registry contains population-based information of 96% of all

pregnancies in The Netherlands. Source data are collected by 95% of midwives, 99% of

gynaecologists and 68% of paediatricians (including 100% of Neonatal Intensive Care

Unit paediatricians).5 From this registry we selected perinatal outcome data of the city of

Rotterdam (587,161 inhabitants in 2009) for the period of 2000-2006 (50,000 singleton

pregnancies). The prevalence of adverse perinatal outcomes was determined for every

neighbourhood using the 4-digit zip codes as recorded in The Netherlands Perinatal Registry,

and municipal neighbourhood boundaries. Adverse perinatal outcome was defined as

the occurrence of either perinatal mortality or perinatal morbidity. Perinatal mortality

was defined as death from 22 weeks of gestational age until 7 days postpartum; perinatal

morbidity was defined as the presence (single or combined) of congenital anomalies

(list defined), small for gestational age (SGA, birthweight below the 10th percentile for

gestational age6), preterm birth (<37th week of gestation) or low Apgar score (< 7, 5 minutes

after birth). The prevalences of adverse perinatal outcomes were illustrated on a map of

Rotterdam making use of separate colours per stratum.

20

URBA

N PERIN

ATAL H

EALTH

INEQ

UA

LITIES

2

Table 2.1 Socio-demographic characteristics of the Rotterdam population (2009, top) and characteristics of the Rotterdam pregnant population along with perinatal outcomes (2000-2006, bottom)

Absolute number Percentage

General population

EthnicityWestern 373,115 64%Non-Western

Surinamese 52,206 9%Antillean 20,261 3%Turkish 46,203 8%Moroccan 38,158 6%Cape Verdean 15,103 3%Non-Western, other 42,115 7%

Socioeconomic characteristicsHaving a job 300,236 51%Dependent on social security 29,504 5%Debt settlement 4,747 1%

Composition of householdLiving alone 139,367 24%Couple without children 133,681 23%Couple with children 216,874 37%One-parent household 80,543 14%Other 16,696 3%

RESULTSTable 2.1 shows recent socio-demographic characteristics of the Rotterdam population in

2009 along with characteristics of the pregnant population from 2000-2006. During the

study period (2000-2006) there were 50,000 singleton pregnancies in the city of Rotterdam.

In almost half of these cases (49%) the mother had a non-Western background; almost

half (48%) of women were primiparous. Teenage pregnancies accounted for 4% of all

pregnancies. The geographic distribution of prevalences of perinatal mortality and perinatal

morbidity is illustrated in figure 2.1. The neighbourhood-specific perinatal mortality rates

varied from 2 to 34 per 1,000 births, for congenital anomalies from 10 to 91 per 1,000 births,

for SGA from 38 to153 per 1,000 births, for preterm birth from 34 to 157 per 1,000 births

and for low Apgar score from 4 to 37 per 1,000 births. The highest mortality rates were

observed in deprived neighbourhoods.

Table 2.1 continues on next page.

21

URBA

N PERIN

ATAL H

EALTH

INEQ

UA

LITIES

2

DISCUSSIONImpressive geographic perinatal health inequalities between neighbourhoods were

observed in the second largest city of The Netherlands. In several neighbourhoods perinatal

mortality and morbidity exceeded national levels observed well before the 1960s. The

highest risks of adverse perinatal outcomes were seen in deprived neighbourhoods,

supporting the theory of neighbourhood context as an individual risk factor for adverse

Table 2.1 Continued

Absolute number Percentage

Pregnant population

Total number of single pregnancies 50,000 100%

Neighbourhoods with <500 births 32 40%Neighbourhoods with 500-1,000 births 27 33%Neighbourhoods with >1,000 births 22 27%

EthnicityWestern 25,544 51%Non-Western 24,456 49%

General characteristicsPrimiparity 24,050 48%Teenage pregnancy 2,231 4%Maternal age >36 years 1,548 3%

Absolute number Per 1,000 births

Perinatal outcomes

Perinatal mortality 600 12Top 5 best neighbourhoods 2 2Top 5 worst neighbourhoods 53 23

Congenital abnormalities 1,333 27Top 5 best neighbourhoods 44 13Top 5 worst neighbourhoods 80 63

Small for gestational age 4,704 94Top 5 best neighbourhoods 92 47Top 5 worst neighbourhoods 378 118

Preterm birth 3,865 77Top 5 best neighbourhoods 84 47Top 5 worst neighbourhoods 204 119

Apgar score <7, 5 minutes after birth 1,221 24Top 5 best neighbourhoods 9 5Top 5 worst neighbourhoods 145 36

22

URBA

N PERIN

ATAL H

EALTH

INEQ

UA

LITIES

2

Figure 2.1 Neighbourhood inequalities in perinatal health in Rotterdam: perinatal mortality rates (top) and adverse perinatal outcome rates (bottom), both per 1000 births (2000-2006); yellow lines indicate highways.

23

URBA

N PERIN

ATAL H

EALTH

INEQ

UA

LITIES

2

perinatal outcomes. This may be due to various reasons (single or combined), e.g., stress

due to unsafety or noise, environmental hazards (petrochemical industry, traffic pollution),

and/or healthcare deficits (restricted healthcare access and insufficient healthcare

performance).3,7,8

Ethnicity is an important risk factor for adverse perinatal outcomes.9,10 However, to under-

stand perinatal health patterns in large cities, urban and socio-economic risk factors should

also be considered.7 In particular, the accumulation of risk factors in urban communities

may explain such perinatal health inequalities as studied in large cohort studies in the cities

of Rotterdam (‘Generation R’)10 and Amsterdam (‘ABCD’)7,9. Medical and non-medical risk

factors may synergistically increase the risk of adverse perinatal outcomes to a far greater

extent than can be explained by their individual etiologic contributions.

Some methodological issues should be discussed. The high number of pregnancies in the

7-year study period minimises the effect of yearly fluctuation of neighbourhood prevalences

of adverse perinatal outcomes. As the risk of perinatal outcomes differs significantly

between multiple and singleton pregnancies, only singleton pregnancies were studied.

The Netherlands Perinatal Registry has a high degree of coverage of all pregnancies in The

Netherlands (96% of all pregnancies are registered), another factor which augments the

validity of our findings.

The usage of 4-digit zip codes allowed a clear illustration of the perinatal health inequalities

between neighbourhoods. The disadvantage of using 4-digit zip codes is the great variety

in neighbourhood size, expressed in geographical size as well as the number of inhabitants.

Consequently, there is a risk of overestimating or underestimating the prevalence of rare

adverse perinatal outcomes in small neighbourhoods. To minimise this effect we aggregated

neighbourhood prevalences into two large groups of outcomes: perinatal mortality and

perinatal morbidity.

To our knowledge, the observed degree of perinatal health inequalities has not been

previously described in other Western cities. Considering the substantial long term health

effects of perinatal morbidity for individuals and generations to come, we feel that these

findings warrant further investigation and also should be given priority on the public health

agenda as it is very likely that similar problems exist in comparable settings.

24

URBA

N PERIN

ATAL H

EALTH

INEQ

UA

LITIES

2

REFERENCES1. Barker DJ. Adult consequences of fetal growth

restriction. Clin Obstet Gynecol. 2006;49:270-83.

2. Moster D, Lie RT, Markestad T. Long-term medical and social consequences of preterm birth. N Engl J Med. 2008;359:262-73.

3. Graaf JP de, Ravelli AC, Wildschut HI, et al. Perinatal outcomes in the four largest cities and in deprived neighbourhoods in The Netherlands. Ned Tijdschr Geneeskd. 2008;152:2734-40.

4. Mohangoo AD, Buitendijk SE, Hukkelhoven CW, et al. Higher perinatal mortality in The Netherlands than in other European countries: the Peristat-II study. Ned Tijdschr Geneeskd. 2008;152:2718-27.

5. Perinatal care in The Netherlands 2006 (in Dutch: Perinatale zorg in Nederland 2006). Utrecht: The Netherlands Perinatal Registry; 2008.

6. Kloosterman GJ. On intrauterine growth. Int J Gynaecol Obstet. 1970;8:895-912.

7. Agyemang C, Vrijkotte TG, Droomers M, van der Wal MF, Bonsel GJ, Stronks K. The effect of neighbourhood income and deprivation on pregnancy outcomes in Amsterdam, The Netherlands. J Epidemiol Community Health. 2009;63:755-60.

8. de Medeiros AP, Gouveia N, Machado RP, et al. Traffic-related air pollution and perinatal mortality: a case-control study. Environ Health Perspect. 2009;117:127-32.

9. Goedhart G, van Eijsden M, van der Wal MF, Bonsel GJ. Ethnic differences in preterm birth and its subtypes: the effect of a cumulative risk profile. BJOG. 2008;115:710-9.

10. Troe EJ, Raat H, Jaddoe VW, et al. Explaining differences in birthweight between ethnic populations. The Generation R Study. BJOG. 2007;114:1557-65.

3

Tijdschrift voor Gezondheidswetenschappen. 2012;90:496-503. (Article in Dutch, translated into English for this thesis)

Pregnancy and birth in Rotterdam: a local health report

J. Poeran, S. Denktaş, E. Birnie, H.J. van der Weg, A.J.J. Voorham, E.A.P. Steegers, G.J. Bonsel

26

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

ABSTRACTWithin The Netherlands, the city of Rotterdam has high rates of perinatal mortality and

morbidity (congenital anomalies, preterm birth, small for gestational age, low Apgar score:

so-called ‘Big4’ morbidities). Following this observation, the municipality of Rotterdam

initiated an elaborate public health initiative to reduce perinatal mortality and morbidity.

This initiative, ‘Ready for a Baby’, is a collaboration between the municipality, the academic

medical centre (Erasmus MC) and the local public health authorities (GGD Rotterdam

Rijnmond). The main target is to decrease perinatal mortality and ‘Big4’ morbidities in

Rotterdam. An important first step was the measuring of baseline perinatal health in the

city of Rotterdam. Data were derived from The Netherlands Perinatal Registry for the period

2000-2007 (n=56,443 singleton pregnancies). We looked at prevalences, absolute and

standardised (for parity, maternal age, ethnicity and socioeconomic status), of perinatal

mortality and ‘Big4’ morbidities within the several boroughs of Rotterdam. Additionally,

we looked at socio-demographics and healthcare related factors. Large differences in

perinatal mortality and morbidity exist between boroughs in the city of Rotterdam, with

most likely different causes per borough. In some boroughs healthcare related factors

and environmental factors play an important role while in other boroughs characteristics

of the pregnant women, e.g., lifestyle or ethnicity are more important causes. These data

are crucial in the development of policies specifically targeted to the several boroughs to

decrease perinatal morbidity and mortality within the city of Rotterdam.

27

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

INTRODUCTIONA recent Dutch study showed women in the four largest cities of The Netherlands to be

at increased risk for perinatal mortality and related perinatal morbidity.1,2 In particular,

Rotterdam has high rates of perinatal mortality and preceding perinatal morbidity, the

most important being congenital anomalies, preterm birth (<37 weeks’ gestation), small for

gestational age (SGA, birthweight <p10) and a so-called low Apgar score, 5 minutes after

birth.3 Due to their importance in perinatal mortality (they precede 85% of perinatal mortality)

we refer to these four precursor morbidities as ‘Big4’. In 2008, following the observations in

the city of Rotterdam, the municipal authorities started an elaborate public health initiative

in collaboration with the academic medical centre (Erasmus MC) and the local public health

authority (GGD Rotterdam Rijnmond): the ‘Ready for a Baby’ programme (in Dutch: Klaar

voor een Kind, see www.klaarvooreenkind.nl). The aim of this 10-year initiative is to reduce

perinatal mortality and perinatal morbidity in the city of Rotterdam to the national level.4,5

The first step in this initiative was to thoroughly record perinatal mortality, morbidity, and

their attributing factors in Rotterdam. These attributing factors can pertain to the pregnant

women (e.g., smoking, folic acid use or ethnicity), their environment (e.g., neighbourhood

social quality or environmental pollution), or they can be related to healthcare (e.g., access

to care or availability of facilities). However, as these factors are not registered as such, we

defined several indicators (which could be derived from existing data) to approximate

the separate role of patient, healthcare and environment related risk factors for adverse

perinatal outcome.3 Depending on the role of these specific indicators, neighbourhood- and

risk factor-specific policy will be possible within the city of Rotterdam.6,7 This approach is

analogous to the national ‘Public Health Forecast’ reports (‘VTV’) by the National Institute for

Public Health and the Environment (‘RIVM’).8 This manuscript describes the used methods

and includes a small selection from the Rotterdam borough reports on perinatal health.

These reports have been discussed in the Rotterdam municipal council in 2011 as well as

the borough councils, and are now publicly available.6

METHODS

The Netherlands Perinatal Registry

Data regarding the indicators (socio-demographics, healthcare related and perinatal

outcome) were derived from The Netherlands Perinatal Registry (www.perinatreg.nl),

28

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

containing complete population-based information of >97% of all pregnancies in The

Netherlands. Source data are routinely collected by 94% of midwives, 99% of gynaecologists

and 68% of paediatricians including 100% of Neonatal Intensive Care Unit paediatricians.

Data were available for 2000-2007; only singleton pregnancies were included.

Indicators

Perinatal outcome

The most important outcome indicators were perinatal morbidity and perinatal mortality.

The former comprised the ‘Big4’ morbidities (congenital anomalies, SGA, preterm birth

and low Apgar score). Congenital anomalies are recorded postpartum and classified

through a standard coding system by organ system (8 categories, 71 subcategories, see

www.perinatreg.nl). Preterm birth is defined as birth before 37 weeks of gestational age;

small for gestational age (SGA) is defined as a birthweight below the 10th percentile for

gestational age. An Apgar score is a 0-10 score based on complexion, pulse rate, reflex

irritability, muscle tone and breathing; a low Apgar score is defined as <7, 5 minutes after

birth. Perinatal mortality is the sum of fetal mortality (intrauterine death from 22 weeks of

gestational age) and early neonatal mortality (death in the first 7 days after birth).

Healthcare related indicators

Healthcare related indicators included in the borough reports refer to: (1) the proportion

of women with a late first antenatal booking visit (after 14 weeks of gestational age), and

(2) the level of care at start of labour (primary care under supervision of a community

midwife, or secondary/tertiary care under supervision of an obstetrician). Women with

a late antenatal booking visit will not fully benefit from healthy lifestyle advices during

pregnancy, concerning an optimal pregnancy outcome. Moreover, due to their gestational

age, these women are not able to participate in first trimester screening for chromosomal

anomalies, mainly Down syndrome. Also, ultrasound determination of gestational age is less

precise after the first trimester; this will result in a less precise due date and the definition

of a possible preterm birth.3,9 The second indicator refers to the unique system of obstetric

care in The Netherlands. This system is characterised by risk-based levels of care: primary

care for low risk pregnancies, provided by independently practicing community midwives,

and secondary/tertiary care for high risk pregnancies provided by obstetricians. Through

risk selection, this distinction between assumed low risk and high risk pregnancies is

made by primary care community midwives. Therefore, the proportion of women starting

labour in primary care (supposedly low risk) is an important healthcare related indicator.

29

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

We used the prevalence of ‘Big4’ as an indicator of high risk pregnancy. Based on the ‘Big4’

prevalence per borough, an estimation can be made on the expected proportion of women

starting labour in primary care and the degree to which high risk pregnancies are detected

and referred, e.g., high ‘Big4’ prevalence with low proportion of women starting labour in

primary care may refer to adequate risk selection.

Demographics

Demographic indicators pertain to ethnicity (Western, non-Western), and maternal age

(<20 years, >35 years). Non-Western women are at increased risk for adverse pregnancy

outcome including perinatal mortality.1 This increased risk also regards pregnant women

outside the range of maternal age 20-35 years.10,11

Other

The other indicators refer to parity (primiparous / multiparous) and the environment of the

pregnant woman (neighbourhood social quality as measured by a composite indicator).

Both indicators are associated with perinatal outcome, including perinatal mortality.1,3

‘Deprived neighbourhood’ was defined as having a ‘Social Index’ score of <6. The ‘Social

Index’ (SI) is a composite 0-10 measure indicating neighbourhood social quality as annually

determined by the Rotterdam Centre for Research and Statistics (see www.cos.rotterdam.

nl). The SI is only available for the city of Rotterdam, but is preferred over the less refined

national ‘yes / no deprived neighbourhood’ variable based on 4-digit zip codes and an

official public list of 40 deprived zip code based neighbourhoods.12,13 Moreover, this national

variable indicates some Rotterdam neighbourhoods as ‘non-deprived’ while their Social

Index score indicates otherwise.

Geographics

In the study period, the city of Rotterdam consisted of 13 boroughs (in March 2010, a 14th

was added, Rozenburg). These boroughs were defined based on 4-digit postal code areas, the

lowest level of aggregation in The Netherlands Perinatal Registry. Next to using crosstabs,

we also illustrated outcomes per borough on a map of Rotterdam, using shades of red to

indicate outcome levels (figure 3.1, the darker the shade the more negative the outcome).

Maps were constructed using ESRI® ArcGIS version 9.3 (Environmental Systems Research

Institute, Inc., USA). Geographic information was obtained from the Rotterdam municipal

services. Due to the availability of data from 8 years (2000-2007) we could avoid too small

numbers of pregnancies per geographic area.

30

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

Standardisation

In comparing indicators between the 13 boroughs, differences in socio-demographics have

to be taken into account as they have an effect on perinatal outcome, e.g., a borough with

a high number of non-Western women, a high number of teenage pregnancies, or women

from neighbourhoods with a low Social Index, will generally have a higher prevalence

of adverse perinatal outcomes. To make an ‘honest’ comparison between the boroughs,

differences in socio-demographics have to be accounted for. Standardisation is a method to

account for this. With ‘direct standardisation’ the ‘expected’ outcome levels per borough are

calculated for the scenario that the concerning borough has the same socio-demographic

composition (by maternal age, parity, Social Index and ethnicity) as the whole city of

Rotterdam. Persisting differences in outcome after standardisation, may be attributable to

Figure 3.1 Perinatal mortality (per 1,000 births) in the 13 boroughs of Rotterdam. Rotterdam average 11.5 per 1,000 births. National average 9.7 per 1,000 births. Names of boroughs are abbreviated: HH Hoek van Holland, OV Overschie, HS Hillegersberg-Schiebroek, AL Prins Alexander, NO Noord, DE Delfshaven, CE Stadscentrum, KR Kralingen-Crooswijk, HO Hoogvliet, PE Pernis, CH Charlois, FE Feijenoord, YS IJsselmonde.

HH

OVHS AL

DE

NOKR

CE

YSFECH

CH

HO

PE

HARBOUR OFROTTERDAM

31

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

other factors (than socio-demographics) such as quality of healthcare or local health policies.

The most important limitation of standardisation is that it implies that the standardisation

factors (in this case maternal age, parity, Social Index and ethnicity) are non-modifiable. In

other words: the effect of these four standardisation factors is disregarded when looking

only at standardised figures. Another important limitation refers to boroughs differing

greatly (compared to the whole city of Rotterdam) in socio-demographic composition, with

accompanying large expected differences in perinatal outcomes. Standardisation might

moderate these significant differences seen in absolute outcomes. Therefore, both crude and

standardised outcomes are important to observe. The technique of direct standardisation

is described in more detail elsewhere.14

Selection from the borough reports

As the borough reports are extensive we included only a selection in this manuscript to

illustrate the most important conclusions. Both crude and standardised rates are included

in the borough reports.6 The selection for this manuscript pertains to:

• crude (absolute, unstandardised) perinatal mortality and ‘Big4’ prevalence per borough

(table 3.1);

• crude perinatal mortality per borough visualised on a map of Rotterdam (figure 3.1);

• crude and standardised indicator rates compared for two boroughs (Hoogvliet and

Overschie, table 3.2). These boroughs were chosen to illustrate that factors attributing to

perinatal mortality (increased in both boroughs) are thought to differ for each borough.

RESULTS

Overview

There were 56,443 singleton pregnancies in the study period (2000-2007) in the city of

Rotterdam. Table 3.1 and figure 3.1 illustrate crude perinatal mortality (per 1,000 births) for

each borough. Additionally, table 3.1 shows crude ‘Big4’ prevalence per borough. Perinatal

mortality varies from 3.4 (Hoek van Holland) per 1,000 births to 21.1 (Pernis) per 1,000 births.

‘Big4’ prevalence varied from 140.6 (Hillegersberg-Schiebroek) per 1,000 births to 205.7

(Charlois) per 1,000 births.

32

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

Rotterdam compared to The Netherlands

Table 3.2 provides an overview of the indicators for The Netherlands, Rotterdam, and two

boroughs (Hoogvliet and Overschie). The first two columns (‘A’ and ‘B’) compare the indicators

for The Netherlands with Rotterdam. Columns 3 to 6 refer to the borough of Hoogvliet;

columns 7 to 10 refer to the borough of Overschie.

The first two columns illustrate a higher perinatal mortality for Rotterdam compared to The

Netherlands (11.5 compared to 9.7 per 1,000 births, respectively). Also, ‘Big4’ prevalence

is higher in Rotterdam (181.1 compared to 152.7 per 1,000 births in The Netherlands).

Rotterdam also has a higher proportion of women with a late antenatal booking visit (36.1%

compared to 20.7% for The Netherlands), and a higher proportion of non-Western women

(48.7% compared to 16.2%).

Comparing boroughs: Hoogvliet

Column 3 (‘C’) from table 3.2 shows crude rates for the borough of Hoogvliet. The proportional

change compared to the crude rates for Rotterdam (‘B’) is shown in the 4th column (‘C-B’).

In particular preterm birth (+38.3%) and low Apgar score (+34.0%) are increased in the

Table 3.1 Perinatal mortality (†) and ‘Big4’ per 1,000 births, for The Netherlands, Rotterdam and the Rotterdam boroughs

Geographic area † ‘Big4’

The Netherlands 9.7 152.7Rotterdam 11.5 181.1

BoroughsCharlois 13.3 205.7Delfshaven 13.3 192.8Feijenoord 11.3 181.1Hillegersberg-Schiebroek 7.6 140.6Hoek van Holland 3.4 181.8Hoogvliet 12.5 203.5IJsselmonde 10.3 184.4Kralingen-Crooswijk 12.0 178.6Noord 8.9 174.3Overschie 16.3 185.7Pernis 21.1 165.7Prins Alexander 12.8 168.6Stadscentrum 13.1 171.3

33

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

Ta

ble

3.2

In

dica

tors

list

ed fo

r tw

o bo

roug

hs, H

oogv

liet a

nd O

vers

chie

. Str

ongl

y de

viat

ing

leve

ls a

re c

olou

red

to e

nhan

ce in

terp

reta

tion*

.

Boro

ugh

of H

oogv

liet

Boro

ugh

of O

vers

chie

NLD

SRD

AM

ABS

CRU

Δ%

STN

STN

Δ%

ABS

ABS

Δ%

STN

STN

Δ%

Indi

cato

rA

BC

C-B

DC-

DE

E-B

FE-

F

Pe

rin

ata

l o

utc

om

e (

pe

r 1

,00

0)

Cong

enita

l ano

mal

ies

27.0

26.4

21.4

-19.

0%25

.8-1

7.1%

27.6

4.8%

27.0

2.2%

Pret

erm

birt

h61

.569

.295

.738

.3%

67.9

41.0

%75

.89.

5%70

.57.

6%Sm

all f

or g

esta

tiona

l age

71.5

93.1

89.9

-3.4

%82

.68.

9%93

.60.

5%91

.12.

6%Lo

w A

pgar

sco

re11

.515

.020

.034

.0%

13.0

54.5

%14

.2-5

.2%

14.6

-2.7

%To

tal '

Big4

' mor

bidi

ties

152.

718

1.1

203.

512

.4%

169.

320

.2%

185.

72.

5%18

0.9

2.7%

Perin

atal

mor

talit

y9.

711

.512

.58.

4%10

.320

.9%

16.3

41.8

%11

.344

.2%

Feta

l mor

talit

y6.

98.

38.

52.

5%7.

316

.6%

13.5

63.1

%8.

166

.5%

Early

neo

nata

l mor

talit

y2.

83.

24.

023

.6%

3.1

31.3

%2.

8-1

2.6%

3.2

-11.

8%

He

alt

hca

re r

ela

ted

(%

)

Ant

enat

al b

ooki

ng a

t >14

wee

ks20

.7%

36.1

%36

.0%

-0.3

%27

.5%

31.0

%23

.8%

-34.

1%32

.6%

-27.

1%St

art l

abou

r in

prim

ary

care

49.0

%47

.6%

20.3

%-5

7.4%

46.1

%-5

6.0%

57.4

%20

.4%

47.0

%22

.1%

Pre

gn

an

t p

op

ula

tio

n (

%)

Non

-Wes

tern

wom

en16

.2%

48.7

%36

.6%

-24.

9%N

/AN

/A36

.8%

-24.

4%N

/AN

/AW

este

rn te

ens

1.2%

2.9%

2.9%

-0.5

%N

/AN

/A1.

7%-4

0.6%

N/A

N/A

Non

-Wes

tern

teen

s4.

3%5.

8%6.

9%18

.7%

N/A

N/A

5.2%

-11.

1%N

/AN

/AW

este

rn 3

5+ y

ears

old

20.4

%22

.0%

17.8

%-1

9.1%

N/A

N/A

22.9

%3.

8%N

/AN

/AN

on-W

este

rn 3

5+ y

ears

old

16.6

%14

.8%

15.1

%2.

3%N

/AN

/A12

.5%

-15.

1%N

/AN

/A

Oth

er

(%)

Prim

ipar

ous

wom

en46

.1%

48.0

%44

.6%

-7.1

%N

/AN

/A49

.2%

2.4%

N/A

N/A

Nei

ghbo

urho

od S

ocia

l Ind

ex <

6.0

N/A

58.3

%0.

0%-1

00.0

%N

/AN

/A49

.7%

-14.

8%N

/AN

/A

 -5

% to

-10%

 m

ore

than

-10%

 +5

% to

+10

% 

+10%

to +

15%

 m

ore

than

+10

%

* Sh

ades

of g

reen

repr

esen

t a p

ositi

ve d

evia

tion;

yel

low

, ora

nge

and

red

repr

esen

t a n

egat

ive

devi

atio

n. C

olou

r cod

ing

is n

ot u

sed

for ‘

the

prop

ortio

n of

wom

en st

artin

g la

bour

in p

rimar

y ca

re’ a

s th

ere

is n

o ne

gativ

e or

pos

itive

judg

men

t pos

sibl

e w

ithou

t tak

ing

into

acc

ount

oth

er fa

ctor

s su

ch a

s an

tena

tal r

isk

load

(‘Bi

g4’).

34

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

borough of Hoogvliet (compared to Rotterdam). On the other hand, congenital anomalies

are decreased (-19.0%). The perinatal mortality rate (12.5 per 1,000 births) is increased

compared to Rotterdam (+8.4%); mainly due to an increase in early neonatal mortality

(+23.6%). Another notable finding is that there is a smaller proportion of women starting

labour in primary care under the supervision of a primary care community midwife (-57.4%).

Also, the Hoogvliet population differs greatly from the Rotterdam population with less

non-Western pregnant women (-24.9%) in Hoogvliet, but an increased proportion of non-

Western pregnant teens (+18.7%). There are no neighbourhoods with a low Social Index

score (<6) in this borough.

Column 5 (‘D’) shows standardised rates for Hoogvliet; column 6 (‘C-D’) shows the comparison

between the crude (‘C’) and standardised (‘D’) rates. Except for congenital anomalies

(-17.1%) all adverse perinatal outcomes have a higher prevalence than expected based on

standardisation, in particular preterm birth (+41.0%), low Apgar score (+54.5%) and early

neonatal mortality (+31.3%). Furthermore, the proportion of women with a late antenatal

first booking is higher (+31.0%) than would be expected based on standardisation. By

contrast, the proportion of women starting labour in primary care is lower (-56.0%) than

expected.

Comparing boroughs: Overschie

Analogous to the rates for Hoogvliet, columns 7 (‘E’) and 8 (‘E-B’) from table 3.2 show crude

rates for the borough of Overschie (column 7) and the proportional difference with the

crude Rotterdam rates (column 8). Crude differences with Rotterdam are relatively small

for separate ‘Big4’ prevalences with a maximum difference of +9.5% for preterm birth.

However, relative perinatal mortality is increased (+41.8%), mainly due to an increased fetal

mortality (+63.1%). The proportion of women with a late antenatal booking visit is lower

(-34.1%) compared to Rotterdam. However, a greater proportion of women start labour in

primary care (+20.4%). Also notable is the decreased proportion of non-Western pregnant

women in Overschie (-24.4%), and also less pregnant women living in a neighbourhood

with a low Social Index (<6, -14,8%). Columns 9 (‘F’) and 10 (‘E-F’) show standardised rates

and their comparison with crude rates, respectively. Standardised ‘Big4’ rates do not differ

much compared to the crude rates for this borough, similar to the crude differences with

Rotterdam as shown in the ‘E-B’ column. Again, perinatal mortality is higher (+44.2%), the

proportion of women with a late antenatal booking is lower (-27.1%), and the proportion

of women starting labour in primary care is higher (+22,1%).

35

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

DISCUSSIONIn Rotterdam, large differences exist in crude and standardised perinatal mortality and ‘Big4’

rates between boroughs. Next to differences in outcome, there are also large differences

in socio-demographics and healthcare related factors. The Rotterdam borough reports

on perinatal health provide possible explanations on borough-specific backgrounds on

increased perinatal mortality or ‘Big4’ prevalence, e.g., maternal or child factors such as

primiparity or having a SGA baby, healthcare related factors such as a suboptimal selection

(and referral) of high risk pregnancies, or environment related factors such as living in a

deprived neighbourhood.

The comparison between the boroughs of Hoogvliet and Overschie showed an increased

perinatal mortality in both boroughs, however, with large differences in ‘Big4’ prevalence,

healthcare related factors and the proportion of non-Western pregnant women. What

particularly stands out is the highly increased perinatal mortality without an expected

increased ‘Big4’ prevalence (as precursors of perinatal mortality) in Overschie. This suggests

a small role of ‘Big4’ morbidities in the increased perinatal mortality shifting the focus to

other causes such as healthcare related factors or environment related factors. The borough

of Hoogvliet has relatively high ‘Big4’ rates accompanied by a strongly increased perinatal

mortality. This suggests a high antenatal risk level, caused by precursors of ‘Big4’, e.g.,

smoking during pregnancy. In addition, the high early neonatal mortality (death after a

live birth) suggests suboptimal healthcare related factors around birth.

The differences between crude and standardised rates indicate different backgrounds

on perinatal health per borough. These differences are more pronounced in the borough

reports on perinatal health as all boroughs are described and compared.6 In some boroughs

healthcare related factors and environment are the main contributors, while in other

boroughs mainly characteristics of the pregnant women are the main contributors (e.g.,

ethnicity or lifestyle factors such as smoking).

Comparable initiatives

‘Urban perinatal health’ is a relatively new field within Obstetrics and Public Health, gaining

a lot of attention in The Netherlands. This is readily illustrated by a comparable urban

perinatal health report for the city of Amsterdam.15 The Amsterdam report also concludes

that effects of risk factors for perinatal mortality differ per borough, advising differential

policies per borough. An important difference, compared to the Rotterdam report, is the

absence of standardisation.

36

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

International initiatives pertaining to urban or inner-city perinatal health also exist.16-18

Even though these initiatives are comparable on some points with the Rotterdam and

Amsterdam initiatives, there are mostly large differences. In the United Kingdom, for

example, there are more (reliable) data available on smoking during pregnancy, one of the

most important lifestyle factors in the context of adverse perinatal outcome.19 Data from

The Netherlands Perinatal Registry underestimate the actual smoking prevalence as this

is poorly registered by the caregivers. In addition, the UK studies use larger geographical

areas, with also outcomes being studied per practice or hospital.16-18 Another important to

mention difference refers to the unique system of obstetric care in The Netherlands which

significantly differs from systems in other Western countries.

Cumulation of risks

An important subject in the UK reports is the connection between adverse perinatal

outcome and social deprivation, a specific characteristic of large urban areas.16-18 Indeed, a

large body on literature exists on the association between social deprivation and adverse

perinatal outcome.20-22 Mostly, the effect of social deprivation on low birthweight / SGA,

preterm birth and perinatal mortality is described.20-22 Also, de Graaf et al. describe an

increased perinatal mortality and ‘Big4’ in socially deprived neighbourhoods in the context

of The Netherlands.1 A particular finding in this study is that specifically Western women

appeared at increased risk in socially deprived neighbourhoods. In Rotterdam, this translated

into a 24% increased risk for adverse perinatal outcome for Western women compared to

non-Western women in socially deprived neighbourhoods.1

Another Dutch study in this context was conducted by Timmermans and colleagues.22

They used data from the Rotterdam ‘Generation R’ cohort and concluded that deprived

neighbourhoods are characterised by a cumulation of risk factors which drives the adverse

perinatal outcomes in these neighbourhoods.22 Numerous risk factors may be responsible

for affecting perinatal health, specific for deprived areas are factors such as feelings of stress

due to unsafety or noise pollution, physical environment (e.g., industry, traffic), or problematic

(access to) healthcare.7,20,22-25 In order to gain insight in the background of urban perinatal

health and specific intervention policies in this context, the full range of urban perinatal

risk factors has to be taken into account; medical as well as non-medical (deprivation,

psychosocial, etc.) risk factors.20,26 Large cohort studies such as ‘Generation R’ in Rotterdam27

and ‘ABCD’ in Amsterdam20,28 have resulted in knowledge on numerous risk factors. Outcomes

of these studies are crucial in understanding perinatal health in the city of Rotterdam.

37

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

The ‘Ready for a Baby’ programme

The large differences in borough profiles suggest borough-specific policies in which

collaborations between caregivers are crucial. Spread out across the country there are

currently numerous regional ‘obstetric collaborations’ between midwives and obstetricians

in the same region. At-risk pregnant women are systematically discussed in these

collaborations. Next to just midwives and obstetricians, it is also crucial to collaborate with

other caregivers in the chain of obstetric healthcare (e.g., youth healthcare physicians,

paediatricians or maternity care professionals).

In the municipal public health programme ‘Ready for a Baby’ experiments in healthcare

are conducted aimed to improve communication and collaboration between caregivers

in the chain of obstetric healthcare (see www.klaarvooreenkind.nl).4,5,29 One of the most

important projects is the so called ‘R4U’ project. ‘R4U’ stands for Rotterdam Reproductive

Risk Reduction; it is a scorecard which scores medical and non-medical (e.g., psychosocial)

risk factors in a standardised way at first antenatal booking visit. This project goes beyond

just screening and subsequently links care pathways to specific risk factors, thus enhancing

collaboration. Other initiatives within the ‘Ready for a Baby’ programme are:

• implementation of and research with a new instrument (‘GyPsy’)29 to screen for psycho-

pathology during pregnancy;

• the construction of within-hospital birth centres (opening in October 2009 and March

2011), a place where women can deliver in a safe home-like primary care setting under

supervision of a primary care community midwife;

• extensive education on preconception care aimed at specific groups at risk, e.g., ethnic

minorities;

• strategies to improve communication in the transfer of care from obstetric care to youth

healthcare.

The effect of these initiatives will become apparent in the coming years by continuous

monitoring of perinatal health in the city of Rotterdam.

Conclusion

Within the city of Rotterdam, large differences in perinatal mortality and morbidity (‘Big4’)

exist between boroughs in the city of Rotterdam, with most likely different causes per

borough. In some boroughs healthcare related factors and environmental factors play

38

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

an important role while in other boroughs characteristics of the pregnant women, e.g.,

lifestyle or ethnicity are more important causes. These data are crucial in the development

of policies specifically targeted to the several boroughs and aimed at decreasing perinatal

morbidity and mortality within the city of Rotterdam.

REFERENCES1. de Graaf JP, Ravelli AC, Wildschut HI, et al. Perinatal

outcomes in the four largest cities and in deprived neighbourhoods in The Netherlands. Ned Tijdschr Geneeskd. 2008;152:2734-40.

2. Steegers EAP, De Graaf JP, Laudy JAM, et al. Recht op een goede start. Medisch Contact. 2008;63:100-1.

3. Bonsel GJ, Birnie E, Denktas S, Poeran J, Steegers EAP. Lijnen in de Perinatale Sterfte, Signale-mentstudie Zwangerschap en Geboorte 2010. Rotterdam: Erasmus MC, 2010. (Dutch report on perinatal research prioritisation).

4. Denktas S, Voorham T, Bonsel GJ, et al. Grootste-delijke perinatale gezondheid. Programmatische aanpak van perinatale sterfte in Rotterdam. TSG, 2009; 87: 199-202.

5. Denktas S, Bonsel GJ, Van der Weg EJ, et al. An Urban Perinatal Health Programme of Strategies to Improve Perinatal Health. Matern Child Health J. 2011;16:1553-8.

6. Poeran J, Birnie E, Denktaş S, Steegers EAP, Bonsel GJ. Perinatale gezondheid in Rotterdam. Rotterdam: Erasmus MC, 2011. (Online: http://www.bds.rotterdam.nl/dsresource?objectid=214373) (Rotterdam borough reports on perinatal health).

7. Poeran J, Denktas S, Birnie E, Bonsel GJ, Steegers EA. Urban perinatal health inequalities. J Matern Fetal Neonatal Med. 2011;24:643-6.

8. Van der Lucht F, Polder JJ. Van gezond naar beter: Kernrapport van de Volksgezondheid Toekomst Verkenning 2010. Rijksinstituut voor Volksgezondheid en Milieu; Bilthoven 2010. (online: http://www.vtv2010.nl/object_binary/o9196_RIVM-01-Kernrapport-Van-gezond-naar-beter-VTV-2010.pdf).

9. Chote AA, Koopmans GT, Redekop WK, et al. Explaining ethnic differences in late antenatal care entry by predisposing, enabling and need factors in The Netherlands. The Generation R Study. Matern Child Health J. 2011;15:689-99.

10. Olausson PM, Cnattingius S, Goldenberg RL. Determinants of poor pregnancy outcomes among teenagers in Sweden. Obstet Gynecol. 1997;89:451-7.

11. Cleary-Goldman J, Malone FD, Vidaver J, et al. Impact of maternal age on obstetric outcome. Obstet Gynecol. 2005;105:983-90.

12. Website Ministry of Housing, Spatial Planning and the Environment. Subject ‘Aandachtswijken’ (on-line: http://www.rijksoverheid.nl/onderwerpen/aandachtswijken?#ref-vrom).

13. Gent, van WPC. Realistic regeneration -housing contexts and social outcomes of neighbourhood interventions in Western European cities-; chapter 4. Thesis 2009, University of Amsterdam. (online available: http://dare.uva.nl/document/153004).

14. Roalfe AK, Holder RL, Wilson S. Standardisation of rates using logistic regression: a comparison with the direct method. BMC Health Serv Res. 2008;8:275.

15. Ravelli AC, Steegers EA, Rijninks-van Driel GC, et al. Perinatale sterfteverschillen in Amsterdam. Ned Tijdschr Geneeskd. 2011;155:A3130.

16. Improving maternity care in London: A framework for developing services. National Health Service: London 2011. (online: http://www.londonhp.nhs.uk/wp-content/uploads/2011/03/Improving-maternity-care-in-London_A-framework-for-developing-services.pdf ).

17. Elsheikh A, Malik A, Gardosi J. West Midlands Perinatal & Infant Mortality 2008/9. West Midlands Perinatal Institute; Birmingham 2010. (online: http://www.perinatal.nhs.uk/pnm/KHD_2008-9.pdf ).

39

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

18. Martin T, Gardosi J. Birmingham and The Black Country: Reducing Perinatal Mortality Project, interim report. National Health Service; Birmingham 2007. (online: http://www.perinatal.nhs.uk/rpnm/B&BC_RPM_Project_Interim_Report.pdf ).

19. Andres RL, Day MC. Perinatal complications associated with maternal tobacco use. Semin Neonatol. 2000;5:231-41.

20. Agyemang C, Vrijkotte TG, Droomers M, van der Wal MF, Bonsel GJ, Stronks K. The effect of neighbourhood income and deprivation on pregnancy outcomes in Amsterdam, The Netherlands. J Epidemiol Community Health. 2009;63:755-60.

21. Auger N, Giraud J, Daniel M. The joint influence of area income, income inequality, and immigrant density on adverse birth outcomes: a population-based study. BMC Public Health. 2009;9:237.

22. Timmermans S, Bonsel GJ, Steegers-Theunissen RP, et al. Individual accumulation of heterogeneous risks explains perinatal inequalities within deprived neighbourhoods. Eur J Epidemiol. 2011;26:165-80.

23. de Medeiros AP, Gouveia N, Machado RP, et al. Traffic-related air pollution and perinatal mortality: a case-control study. Environ Health Perspect. 2009;117:127-32.

24. Hartikainen AL, Sorri M, Anttonen H, Tuimala R, Laara E. Effect of occupational noise on the course and outcome of pregnancy. Scand J Work Environ Health. 1994;20:444-50.

25. Wilhelm M, Ritz B. Residential proximity to traf-fic and adverse birth outcomes in Los Angeles county, California, 1994-1996. Environ Health Perspect. 2003;111:207-16.

26. Kramer MS, Seguin L, Lydon J, Goulet L. Socio-economic disparities in pregnancy outcome: why do the poor fare so poorly? Paediatr Perinat Epidemiol. 2000;14:194-210.

27. Troe EJ, Raat H, Jaddoe VW, et al. Explaining differ-ences in birthweight between ethnic populations. The Generation R Study. BJOG. 2007;114:1557-65.

28. Goedhart G, van Eijsden M, van der Wal MF, Bonsel GJ. Ethnic differences in preterm birth and its subtypes: the effect of a cumulative risk profile. BJOG. 2008;115:710-9.

29. Quispel C, Schneider TA, Bonsel GJ, Lambregtse-van den Berg MP. An innovative screen-and-advice model for psychopathology and psychosocial problems among urban pregnant women: an exploratory study. J Psychosom Obstet Gynaecol. 2012;33:7-14.

40

PREGN

AN

CY AN

D BIRTH

IN RO

TTERDA

M: A

LOCA

L HEA

LTH REPO

RT

3

4

Social Science and Medicine. 2013;83:42-9.

Social deprivation and adverse perinatal outcomes among Western and non-Western pregnant women

in a Dutch urban population

J. Poeran, A.F.G. Maas, E. Birnie, S. Denktaş, E.A.P. Steegers, G.J. Bonsel

42

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

ABSTRACTSocial deprivation is considered a key factor in adverse perinatal outcomes. Rotterdam,

the second largest city in The Netherlands, has large inequalities in perinatal health and a

high number of deprived neighbourhoods. Social deprivation is measured here through

a composite variable: ‘Social Index’ (SI). We studied the impact of the SI (2008-2009; 5

categories) in terms of perinatal mortality, congenital anomalies, preterm birth, small

for gestational age (SGA) and low 5-minute Apgar score as registered in The Netherlands

Perinatal Registry (Rotterdam 2000-2007, n=56,443 singleton pregnancies). We applied

ethnic dichotomisation as Western (European/North-American/Australian) vs. Non-Western

(all others) ethnicity was expected to interact with the impact of SI. Tests for trend and

multilevel regression analysis were applied. Gradually decreasing prevalence of adverse

perinatal outcomes was observed in Western women from the lowest SI category (low social

quality) to the highest SI category (high social quality). In Western women the low-high SI

gradient for prevalence of spontaneous preterm birth (per 1,000) changed from 57.2 to 34.1,

for iatrogenic preterm birth from 35.2 to19.0, for SGA from 119.6 to 59.4, for low Apgar score

from 10.9 to 8.2, and for perinatal mortality from 14.9 to 7.6. These trends were statistically

confirmed by Chi2-tests for trend (p<0.001). For non-Western women such trends were

absent. These strong effects for Western women were confirmed by significant odds ratios

for almost all adverse perinatal outcomes estimated from multilevel regression analysis.

We conclude social deprivation to play a different role among Western vs. non-Western

women. Our results suggest that improvements in social quality may improve perinatal

outcomes in Western women, but alternative approaches may be necessary for non-Western

groups. Suggested explanations for non-Western ‘migrant’ groups include the presence of

‘protective’ effects through knowledge systems or intrinsic resilience. Implications concern

both general and targeted policies.

43

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

INTRODUCTIONIn The Netherlands perinatal mortality exceeds the European average despite a high standard

of mother and child healthcare with free access.1 Perinatal health in the larger cities is even

worse, with the highest rates of perinatal mortality and morbidity being observed in deprived

neighbourhoods.2 The high prevalence of ethnic minority groups and disseminated social

deprivation in urban areas are generally put forward as key aetiologic factors.2-6 Social

deprivation is a very broad term and can be defined as ‘reduction or prevention of culturally

normal interaction with the rest of society’. Indeed, aspects of social deprivation such as

material poverty and lack of social cohesion are both related to ill health, and also strongly

connected; the combined reinforcing presence of these factors might be particularly important

for perinatal ill health.7-9 Numerous studies have shown ethnicity and social deprivation to be

strongly related to adverse perinatal outcomes such as preterm birth and small for gestational

age.3,4,10-14 However, many recent studies have been conducted in the United States and Canada

where ethnic minorities differ considerably from those in Europe and, more specifically, The

Netherlands.2-6,11,13,15,16 In the United States the majority of ethnic minorities either comprise

African-Americans or Hispanics; in Europe they mainly originate from former colonies (for

example in the United Kingdom or The Netherlands) or they result from the 1960’s labour

migration from countries such as Turkey or Morocco (for example in Germany and France,

respectively). Findings from these studies do not necessarily apply to European countries.

Another motivation for our study pertains to findings from a recent Dutch study

which showed Western (European/North-American/Australian) women in deprived

neighbourhoods to have an increased risk of adverse perinatal outcomes as opposed to

non-Western women.2

Rotterdam, the second largest city of The Netherlands, has the highest proportion of non-

Western inhabitants as well as the highest number of deprived neighbourhoods, and the

highest rate of adverse perinatal outcomes, creating a suitable population in which to study

the effect of social deprivation on perinatal outcomes.2 In continuation of previous work, we

investigated the background and the association of social deprivation with adverse perinatal

outcomes, for Western and non-Western women separately, as we hypothesise differential

effects. We used a composite variable, the so called ‘Social Index’ (SI) as deprivation indicator

at the neighbourhood level in the city of Rotterdam. As social deprivation is considered

an important metric of neighbourhood quality policy makers have created the SI and its

underlying domains to measure this. It is used to measure the effectiveness of efforts to

reduce area-based social deprivation. The SI conceptually resembles the less detailed

44

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

Scottish Carstairs index.17 We use the unaltered SI values to facilitate communication of

study results to policy-makers.

METHODS

Outcome data

Data from all single pregnancies in Rotterdam over the period 2000-2007 were derived from

The Netherlands Perinatal Registry. This registry contains population-based information of

97% of all pregnancies in The Netherlands.18 Source data are collected by 94% of midwives,

99% of gynaecologists and 68% of paediatricians, including 100% of Neonatal Intensive

Care Unit paediatricians.18 The mission of The Netherlands Perinatal Registry is to improve

the quality of healthcare by giving insight into the perinatal care process and outcomes

(see also www.perinatreg.nl).

The Netherlands Perinatal Registry provided individual-level information on adverse

perinatal outcomes, along with the four-digit zip codes of the mothers’ places of residence.

The number of pregnancies per zip code area (neighbourhood) ranged from 127 to 2,611.

Adverse perinatal outcome was defined as the occurrence of either perinatal mortality or

perinatal morbidity. Perinatal mortality was defined as death from 22 weeks of gestational

age until 7 days postpartum. We also defined four outcome variables of perinatal morbidity:

congenital anomalies (list defined), small for gestational age (SGA, birthweight below

the 10th percentile for gestational age19, preterm birth (<37th week of gestation) or low

5-minute Apgar-score (<7). As previously described4, preterm birth was subdivided into

spontaneous and iatrogenic preterm birth, the latter being defined as birth <37th week

of gestation and an elective caesarean section or induction of labour. The remainder of

births <37th week of gestation were classified as spontaneous preterm birth. We assume

iatrogenic preterm birth and low Apgar score to be related to both maternal factors as well

as peripartum healthcare factors, whereas the remaining morbidity conditions are assumed

to be primarily related to individual (maternal) characteristics.

Social Index

Neighbourhood social quality was assessed making use of a combined variable, the so-called

‘Social Index’ (SI). This index was created in 2008 and is calculated annually for the Rotterdam

municipal authorities by the Centre for Research and Statistics Rotterdam (COS-Rotterdam,

45

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

www.cos.nl) as a policy measure. All underlying data are empirical. The SI is a composite

multidimensional variable indicating neighbourhood social quality on a 1-10 scale. As is

depicted in figure 4.1 the SI sum score combines scores from 4 ‘domains’: (1) (personal)

capacities, (2) living environment, (3) participation and (4) neighbourhood commitment

(‘social cohesion’). In turn, these ‘domain’ scores are a sum of ‘item’ scores which are based

on a combination of registration data and questionnaire data. The questionnaire data

are obtained from respondents from a random sample of the Rotterdam population. Per

neighbourhood 900 persons were sampled, proportionally stratified for age group, sex and

ethnicity. The number of inhabitants per neighbourhood ranged from 1,579 to 21,200. The

aim was to have 175 respondents (who completed the questionnaire) per neighbourhood.

The initial format was an online questionnaire; if response was not sufficient, respondents

were approached with a paper and pencil questionnaire or by telephone. Overall, response

was 52% online, 21% by paper and pencil, and 27% by telephone.

Figure 4.1 Composition of the SI: four SI ‘domains’ with corresponding ‘items’ extracted from questionnaire data (*) and from registration data (†).

SOCIAL INDEX

(Personal) Capacities Living Environment Participation NeighbourhoodCommitment

EDUCATION-school drop-outs†

-starting qualification / educational level†

(PERCEIVED) HEALTH-perceived health*-filed reports at local(public) health initiatives†

-filed reports of domestic violence†

INCOME-low income households*-social welfare receivers†

-insufficient incomehouseholds*

PROFICIENCY-Dutch proficiency*

DISCRIMINATION-perceived discrimination*-perceived social inter-course between ethnicgroups*

HOUSING QUALITY-overcrowded housing†

-satisfactory housing*

NEIGHBOURHOOD FACILITIES-satisfactoryneighbourhood (health/social) facilities*-knowledge ofneighbourhood socialfacilities*

NUISANCES-neighbourhood pollution*-neighbourhoodvandalism*-neighbourhoodnuisances from speedingcars, traffic noise,rumours, drugs andhang-around youths*-perceived safety*

SCHOOL, WORK AND YOUTHS-job applicants†

-school going youth†

SOCIAL CONTACTS-contacts with family andfriends*-social isolation*

SOCIALPARTICIPATION-sports, social and cultural activities, going out*

DEDICATION TO SOCIETAL ACTIVITIES-dedication to informalcare*-dedication to voluntarywork*-active participation to neighbourhood liveability*

NEIGHBOURHOOD MIGRATION-neighbourhood migrationrate† -neighbourhood newresident rate†-length of residency†

PERCEIVED NEIGH-BOURHOOD BINDING-perceived neighbour-hood binding*-dedication to neighbour-hood (activities)*-trust in local authorities*

46

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

The SI and its underlying domains are considered important metrics of the effectiveness of

policy makers’ efforts to reduce area-based social deprivation. More in-depth information

on the Social Index and its construction is provided in the online appendix file. In this study,

we used the average SI for 2008 and 2009 as a proxy for neighbourhood social quality,

both as a continuous measure and as an ordinal measure (5 categories) with the following

‘COS-Rotterdam’ thresholds:

<3.9 highest social deprivation (I);

3.9-4.9 problematic social deprivation (II);

5.0-5.9 moderate social deprivation (III);

6.0-7.0 adequate social quality (IV);

>7.0 high social quality (V).

Ethnicity

Dutch law does not permit the routine utilisation or registration of ethnicity data. As yet,

The Netherlands Perinatal Registry is exempt from this restriction. The ethnic classification

in this professional-based registry recognises seven possible categories: Western Dutch,

Western other (including women from other European countries, Australia and the US),

Mediterranean, (East) Asian, African, South Asian, or other non-Western. The African and

South Asian group mainly comprises women from the former Dutch colonies Suriname and

The Netherlands Antilles. The group of East Asian women mainly originate from Indonesia,

also a former Dutch colony.

Classification of clients is done by the healthcare professional with an absence of strict

coding rules in applying the category labels. Current classification is therefore crudely based

on a mixture of self-declared ethnicity, race and known country of birth of the woman or

her parents. The registry does not contain information on first/second generation migrants,

nor on their length of stay in Rotterdam or The Netherlands.

Therefore, the current system as devised by the professional organisations may introduce

some classification error, in particular within ‘migrant’ (non-Western) categories. We

dichotomised ethnic groups as Western vs. non-Western for two reasons: (1) to compare

our results with previous studies, and (2) in view of the hypothesis generating nature of

our study which may lead to more detailed investigation of specific ethnic groups in future

studies. The first two classes of the original classification were combined into Western and

the remainder into non-Western women.2

47

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

Analysis

Individual-level data on pregnancy outcomes as registered in The Netherlands Perinatal

Registry were linked to the 2008-2009 neighbourhood SI making use of 4-digit zip codes.

Simple crosstabs were created illustrating the proportion of Western and non-Western

women, mean age, and prevalence rates of adverse perinatal outcomes for each SI category.

Mean age and adverse outcome were separately illustrated for Western and non-Western

pregnant women. One-way tests for trend were carried out for the association of ethnicity,

age and adverse perinatal outcomes and the SI categories (ANOVA for age, Chi-square for

dichotomous variables).

Analogous to the above, we also compared average SI domain scores for Western women

with scores from non-Western women in each SI category. A one-way ANOVA test was used

to test for the difference in mean SI domain scores.

Multilevel logistic regression models were used to estimate adjusted odds ratios (aOR) to

further asses the association between the 2008-2009 mean neighbourhood SI (continuous

measure) and adverse perinatal outcomes (dichotomous outcomes) for Western and non-

Western women. We specified hierarchical random-intercept models that allow for the

incorporation of both individual-level and neighbourhood-level characteristics, as well as

the adjustment for clustering of individuals within their neighbourhoods. SAS version 9.2

(SAS Institute Inc., Cary, NC) was used to run the random intercept multilevel models with

the GLIMMIX procedure. Crude and aORs (for parity; primiparous/multiparous) are reported.

The significance was set at alpha=0.05, two tailed.

RESULTSA total of 56,443 singleton pregnancies were analysed. Characteristics of the study

population are shown in table 4.1.

Patient characteristics per SI category

The association of ethnicity and mean maternal age with SI category is described in table

4.2 (upper part). There are no neighbourhoods in Rotterdam in category I (SI <3.9). The

proportions of Western women and non-Western women respectively increase (from 24.0

to 89.5%) and decrease (from 76.0 to 10.5%) with increasing SI (a higher SI stands for a

neighbourhood with higher social quality). Mean maternal age increases for both Western

48

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

Table 4.1 Characteristics of the study population and perinatal outcomes (Rotterdam, 2000-2007)

  N Percentage

No. of singleton pregnancies during study period 56,443 100.0%

Paritya

Primiparous 27,105 48.0%Multiparous 29,333 52.0%

Gestational agea

< 37 weeks 3,900 7.0%37-42 weeks 51,591 92.9%> 42 weeks 71 0.1%

Maternal agea

≤ 19 years 2,445 4.3%20-24 years 10,378 18.4%25-29 years 15,800 28.0%30-34 years 17,368 30.8%≥ 35 years 10,437 18.5%

EthnicityWestern 28,972 51.3%Non-Western 27,471 48.7%

Social Index>3.9 0 0.0%3.9-4.9 6,138 10.9%5.0-5.9 26,760 47.4%6.0-7.0 16,611 29.4%>7.0 6,934 12.3%

   Absolute number Per 1,000 births

Perinatal outcomesb

Congenital anomalies 1,489 26.4Preterm birth 3,908 69.3Small for gestational age 5,255 93.1Apgar score <7 (5 minutes after birth) 844 15.0

Any Big4b 10,222 181.2

Fetal mortalityc 466 8.3

Early neonatal mortalityd 183 3.2

Perinatal mortalitye 649 11.5a Totals do not add up due to missing values (parity 5 missings, gestational age 881 missings, maternal age 15 missings).b Number and proportion of pregnant women with at least one ‘Big4’ morbidity.c From 22 weeks of gestational age.d 0-7 days postpartum.e Total of fetal, intrapartum and neonatal mortality.

49

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

Ta

ble

4.2

M

ean

age,

eth

nici

ty a

nd p

reva

lenc

e ra

tes

of a

dver

se p

erin

atal

out

com

es fo

r Wes

tern

and

non

-Wes

tern

wom

en p

er S

I cat

egor

y

 Et

hnic

ityTo

tal

SI c

ateg

ory

pa

III

IIIIV

V 

 n=

56,4

43n=

0n=

6,13

8n=

26,7

60n=

16,6

11n=

6,93

4 

Preg

nanc

ies

(%)

Wes

tern

51.3

024

.038

.067

.089

.5<0

.001

Non

-Wes

tern

48.7

076

.062

.023

.010

.5

Mea

n ag

e (y

ears

)W

este

rn30

.30

27.7

29.7

30.3

32.1

<0.0

01N

on-W

este

rn28

.00

27.4

28.0

28.3

30.0

<0.0

01

Cong

enita

l ano

mal

ies

(per

1,0

00)

Wes

tern

25.9

023

.826

.423

.530

.00.

249

Non

-Wes

tern

26.9

020

.828

.826

.424

.60.

219

Pret

erm

birt

h (p

er 1

,000

)Sp

onta

neou

sW

este

rn44

.50

57.2

49.1

44.4

34.1

<0.0

01N

on-W

este

rn45

.20

46.6

44.8

45.9

42.0

0.75

7Ia

trog

enic

Wes

tern

26.4

035

.225

.330

.319

.00.

009

Non

-Wes

tern

24.5

023

.623

.427

.533

.60.

056

Smal

l for

ges

tatio

nal a

ge (p

er 1

,000

)W

este

rn84

.20

119.

699

.179

.959

.4<0

.001

Non

-Wes

tern

103.

40

103.

310

2.5

106.

010

6.1

0.59

3

Low

Apg

ar s

core

(per

1,0

00)

Wes

tern

12.5

010

.913

.514

.38.

20.

050

Non

-Wes

tern

17.5

018

.618

.015

.513

.70.

139

Perin

atal

mor

talit

y (p

er 1

,000

)W

este

rn10

.30

14.9

11.1

10.5

7.6

0.00

8N

on-W

este

rn12

.70

12.0

13.3

12.4

6.8

0.57

6

 a Chi

-squ

are

test

for t

rend

was

use

d fo

r eth

nici

ty a

nd p

reva

lenc

e ra

tes

of a

dver

se o

utco

mes

. For

the

mea

n ag

e a

one-

way

AN

OVA

was

use

d. 

50

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

and non-Western women with increasing SI (27.7-32.1 and 27.4-30.0 years, respectively).

Both mean age (for Western and non-Western women) and ethnicity show significant

p-values (p<0.001) when testing for trend.

Outcomes in Western women per SI category

The lower part of table 4.2 shows a decreasing trend for rates of spontaneous and iatrogenic

preterm birth (range: 57.2-34.1 and 35.2-19.0 per 1,000, respectively), small for gestational

age (119.6-59.4 per 1,000) and perinatal mortality (14.9-7.6 per 1,000) with better SI values.

This is also verified by the significant p-values when testing for trend (range: <0.001-0.009).

When comparing SI category II with category V, the congenital anomaly rate increases (23.8-

30.0 per 1,000), and the low Apgar score rate decreases (10.9-8.2 per 1,000).

Outcomes in non-Western women per SI category

When comparing SI category II with category V, congenital anomalies, iatrogenic preterm

birth and small for gestational age increase (from 20.8 to 24.6, 23.6-33.6 and 103.3-106.1

per 1,000, respectively). Conversely, rates of spontaneous preterm birth, low Apgar score

and perinatal mortality decrease (46.6-42.0, 18.6-13.7 and 12.0-6.8 per 1,000, respectively).

Trends, however, were not statistically significant.

SI domains

Table 4.3 depicts mean SI domain scores per SI category for Western and non-Western women.

Western women almost always have significantly higher mean SI domain scores compared

with non-Western women, except for SI domains ‘participation’ and ‘neighbourhood

commitment’ in SI category II. The latter being the only domain score for which non-Western

women have a significantly higher score compared with Western women.

Multilevel logistic regression models

Table 4.4 shows the crude and adjusted odds ratio (OR and aOR adjusted for parity) of

the impact of the mean 2008-2009 neighbourhood SI (on a continuous 1-10 scale) on

adverse perinatal outcomes, for Western and non-Western women. For Western as well as

non-Western women congenital anomalies show non-significant effects of SI. For Western

women significant negative associations with SI were shown for spontaneous and iatrogenic

preterm birth, small for gestational age, low Apgar score and perinatal mortality. For non-

51

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

Western women a positive association was demonstrated for iatrogenic preterm birth and

a negative association for low Apgar score. The largest protective effect of higher SI was

seen in Western women for small for gestational age (aOR 0.78; CI 0.73-0.83).

Table 4.3 Mean SI domain scores per SI category for Western and non-Western women

    SI category

I II III IV VDomain Ethnicity n=0 n=6,138 n=26,760 n=16,611 n=6,934

Capacities Western 0 3.94a 4.86a 6.44a 8.13a

Non-Western 0 3.93a 4.69a 6.33a 8.03a

Living environment Western 0 5.13a 5.88a 6.71a 7.61a

Non-Western 0 5.10a 5.78a 6.64a 7.55a

Participation Western 0 5.03 5.77a 6.34a 7.37a

Non-Western 0 5.04 5.72a 6.29a 7.32a

Neighbourhood commitment Western 0 4.85a 5.71b 6.62a 7.41a

Non-Western 0 4.93a 5.72b 6.56a 7.32a

a Western vs. non-Western women, ANOVA p<0.001.b Western vs. non-Western women, ANOVA p<0.05. 

Table 4.4 Multilevel hierarchical random-intercept logistic regression models: crude (OR) as well as adjusted (aOR, adjusted for parity) odds ratios including 95% confidence intervals (CI) of the associations between mean 2008-2009 neighbourhood SI and adverse perinatal outcomes for Western and non-Western women

   Ethnicity ORa (CIc) aORb (CIc)

Congenital anomalies Western 1.02 (0.90-1.16) 1.02 (0.90-1.17)Non-Western 1.02 (0.91-1.15) 1.02 (0.91-1.15)

Preterm birthSpontaneous Western 0.86 (0.80-0.92)d 0.87 (0.80-0.93)d

Non-Western 1.01 (0.92-1.11) 1.00 (0.91-1.10)Iatrogenic Western 0.88 (0.78-0.98)e 0.88 (0.78-0.99)e

Non-Western 1.15 (1.03-1.28)e 1.15 (1.03-1.28)e

Small for gestational age Western 0.77 (0.72-0.82)d 0.78 (0.73-0.83)d

Non-Western 1.02 (0.96-1.08) 1.01 (0.95-1.07)

Low Apgar score Western 0.83 (0.73-0.95)e 0.84 (0.74-0.96)e

Non-Western 0.87 (0.75-1.00) 0.86 (0.74-1.00)e

Perinatal mortality Western 0.83 (0.72-0.96)e 0.84 (0.72-0.97)e

Non-Western 0.96 (0.82-1.13) 0.96 (0.82-1.12)a OR: crude odds ratiob aOR: adjusted odds ratio (for parity)c CI: confidence intervald p-value <0.001e p-value <0.05

52

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

DISCUSSION

Principal fi ndings

This study is one of the few European studies to address the effect of a combined social

deprivation measure on adverse perinatal outcomes and applying multilevel modelling.20 In

the large, multi-ethnic city of Rotterdam the impact of social deprivation on adverse perinatal

outcomes plays a key role with a striking ethnicity-related effect. In the most deprived

neighbourhoods, perinatal outcomes were universally poor with a tendency for even worse

figures for Western women compared to non-Western women. With decreasing deprivation, a

strong gradient for improvement exists for almost all perinatal outcomes in Western women.

For every point increase of the Social Index (higher is better), adjusted perinatal mortality

in Western women shows a 16% decrease. As a similar gradient in perinatal outcomes is

absent in non-Western women, an ethnicity related perinatal outcome gap emerges, with

impressive perinatal health inequalities in neighbourhoods with the lowest deprivation.

Other studies

Our finding that social deprivation has a stronger negative effect on Western women than

non-Western women has been sparsely described in the literature.13,21 Fang et al. used New

York City birth records to show that foreign born women of African and Caribbean descent

had more favourable birth outcomes compared with that for whites, particularly in the

poorest neighbourhoods.21 O’Campo et al. studied the effect sizes for non-Hispanic whites

and non-Hispanic blacks for the relation between neighbourhood deprivation and preterm

birth.13 For whites, the effect size was larger compared to blacks indicating that this group

appears to be less affected by neighbourhood deprivation than whites.13 Furthermore, one

previous study by De Graaf et al., conducted in The Netherlands also supports the current

findings.2 They found that the effect of living in deprived neighbourhoods, as compared

to living elsewhere, was most prominent among Western women with an increased risk of

adverse perinatal outcomes of 24%.2

Possible underlying mechanisms

It is challenging to interpret the observed ethnicity-related effect by one underlying

mechanism. A first explanation takes into account the density of migrant subgroups in the

neighbourhood population.22 Previous studies have suggested that the risk of poor health

53

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

outcomes for a minority individual is inversely related to the density of his or her racial/ethnic

group in the local community, i.e., greater numbers of migrants/non-Western populations in

deprived neighbourhoods, and that living in neighbourhoods with a high migrant density

may be beneficial specifically for migrant (here non-Western) women, such that fewer

negative effects of deprivation in these neighbourhoods may be experienced. This hypothesis

has been validated for various adverse health outcomes ranging from psychiatric conditions

to rates of heart disease.7,8,23,24 The presumed mechanisms include health protection through

increased participation in social networks or knowledge systems, and a more extensive

repertoire of positive coping behaviours.7,8,23,24 When applied to our study context, these

protective effects may result in less feelings of stress during pregnancy and may stimulate

healthy behaviour. In Rotterdam, this could be valid for the non-Western women in deprived

areas (lowest SI) as they generally tend to have better outcomes than Western women in these

areas. However, we also observed a lack of improvement in adverse perinatal outcomes with

increasing SI for non-Western women. This ‘net neutral effect’ may be (partly) explained by

the combination of a decrease in potentially protective migrant/non-Western density when

SI and its positive effects increase. A second explanation may rest on lifestyle epidemiology

concerning factors such as smoking which is known to have a profound effect on perinatal

outcome.25 Previous studies have shown that smoking behaviour differs among ethnic and

socio-economic groups.26 Our data could reflect healthier lifestyle behaviour for Western

women with increasing SI, this gradient may be absent in non-Western women.

Another explanation of our study results pertains to healthcare related factors. This may

pertain to general thresholds to healthcare or density of facilities which is lower in deprived

areas. Western women might profit more from this effect due to better health literacy and

informal networks.9 We assumed iatrogenic preterm birth and low Apgar score to be partially

related to peripartum healthcare factors. However, different patterns were observed for

low Apgar score and iatrogenic preterm birth in Western and non-Western women. Thus,

it is unclear to what extent healthcare related factors have affected the observed ethnic

disparities. The perinatal healthcare process still needs careful analysis as little is known on

its dynamics in a multicultural setting.27

A sociobiological explanation of our study results may involve the presence of epigenetic

effects in some groups of non-Western women, particularly those of African descent, that

limit the potential for positive effects on perinatal outcomes when SI increases.28,29 This

phenomenon has been demonstrated by Collins et al. in a three generation linked dataset of

African Americans in Illinois.30 Their study showed rates of low birthweight to be associated

with worsening maternal grandmothers’ residential environments during her pregnancy

54

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

with her daughter that years later delivered a low birthweight neonate. This association

was independent of the living conditions of the daughter during her pregnancy with the

infant with low birthweight.30 Further research is needed to investigate if this phenomenon

among African Americans also applies to other non-Western populations.

Finally, instead of the above so-called ‘causation’ options, one may think of ‘selection’. Western

and non-Western women may differ in their motive to move to a deprived neighbourhood

or differ in the potential for ‘upward mobility’. As previously described, ethnic minorities

tend to cluster, usually in poor urban areas. This ‘ethnic concentration’ may be maintained by

both its residents and newly arriving migrants because of the amenities and organisations

feeding on this ethnic concentration. That is why residents stay in these neighbourhoods

even when they have the financial means to move out.31,32 Western women in deprived

neighbourhoods are thought to represent a negative selection as the above mechanism

may not apply to them.31,32 The latter is supported by our finding that the only SI domain

for which Western women have a significantly lower score than non-Western women is for

‘neighbourhood commitment’ (social cohesion) in the lowest SI category in Rotterdam, i.e.,

Western women are thought to have less commitment to these neighbourhoods.32

Strengths and weaknesses

The major strengths of the current study include the use of a validated registry with a high

coverage over a long period of time; The Netherlands Perinatal Registry, currently 2000-

2007.18 As social quality and social deprivation are multidimensional concepts, the usage

of a neighbourhood-specific combined social deprivation variable adds strength to the

estimates and, in case of the Social Index, it also allows to draw conclusions regarding a

more specific dimension of social deprivation.33 Neighbourhood-specific social deprivation

indices are generally constructed with factor or principal component analysis.16,33 Such an

approach rests on selecting the most efficient items to represent 1 or more underlying

constructs. The SI intentionally was composed by municipal policymakers, selecting items

which in their view represented best observable and direct interpretable information

on social deprivation; herewith, statistical overlap was accepted to allow for political

communication. Hence, the Social Index was designed for policy monitoring purposes as

an outcome in the first place, rather than for research purposes as a determinant. What also

adds strength to our findings is that initial results were cross-validate d by additional analyses

of the effect of neighbourhood house price quintiles on adverse perinatal outcomes (data

available in the online appendix file). A particular strength of the current study pertains to

55

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

the differentiation between iatrogenic and spontaneous preterm birth as they may differ in

aetiological or causal pathways.34 The majority of previous retrospective studies on the effect

of social deprivation and preterm birth do not take this difference into account.10,12,15,20,34,35

The limitations associated with this study should also be noted. Firstly, area based indices

such as the SI may not correspond to individual socioeconomic status (‘ecological fallacy’)

and do not reflect heterogeneity among the individuals within a neighbourhood. A

possible way to triangulate would be to interview a sample of the women. However, the

strict national privacy regulations which apply to The Netherlands Perinatal Registry,

do not allow identification of individuals for subsequent active approach. Multilevel

modelling to estimate effects based on a neighbourhood-level measure may partially

bypass this problem. The advantage of this approach is that it allows for the incorporation

of both individual-level and neighbourhood-level characteristics. Moreover, it adjusts for

clustering of individuals within neighbourhoods, thereby increasing the validity of the

effect sizes compared to conventional logistic regression analysis.36 As parity was one of

the few individual-level factors we could adjust for, and as we did not have information

on individual-level socioeconomic status, we could not fully benefit from the theoretical

advantages of multilevel modelling over conventional logistic regression. An advantage

of the usage of neighbourhood specific indices is that they provide complementary

information to individual measures of social quality and may be particularly useful in studies

concerning pregnancy outcomes. Determination of social class is known to be problematic

in such studies as not only the mother’s socioeconomic characteristics have to be taken

into account, but also the father’s.37 Additionally, the use of area based deprivation indices

is well supported in the literature.3,10-13,15,16,20,37

Another limitation refers to the use of the mean 2008-2009 SI in assessing the effect on

perinatal outcomes from 2000-2007. This discrepancy was due to the unavailability of data.

Consequently, our analysis does not take into account the variability of neighbourhood SI

between 2000 and 2007. The variability between 2008 and 2009 was small (data available

in the online appendix file). Moreover, it is generally expected to take a long period of time

for possible changes in neighbourhood social quality to take effect (data available in the

online appendix file). Therefore, the discrepancy in the years of available data between

the exposure variable and outcome variables is expected to play a marginal role. Also, the

current study lacks data on lifestyle, in particular smoking, one of the main risk factors for

adverse perinatal outcomes.25 This could have resulted in possible overestimation of the

impact of SI on adverse pregnancy outcomes. However, previous studies have shown that

socioeconomic status remains significantly associated with adverse perinatal outcomes

56

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

such as preterm birth, even after taking smoking into account.13,20 Finally, one could argue

against the grouping of all non-Western women as this is a very heterogeneous group,

disregarding among others any differences in races, cultures, first or second generations, or

migrants’ length of stay. However, the main objective of our study follows from a previous

study which suggested that specifically Western women were at risk for adverse pregnancy

outcomes in deprived areas.2 To further study and reconfirm this effect, we chose to group

non-Western women as we were specifically interested in Western women.

Possible implications

The SI is associated with adverse perinatal outcomes to the extent that both general and

targeted policies seem relevant. In the context of social deprivation, social policy should

continue to aim to serve Western as well as non-Western women equally. General policies,

in particular preventive policies, show higher average benefits (compared to targeted

policies) if applied on the population level. However, they do so while increasing health

gaps as their effects rely on competences and resources which are unequally distributed,

this is called the ‘prevention paradox’.38 Therefore, complementary priority programs are

justified, dedicated to specific ‘worst off’ subgroups.

The observed increased risk of adverse perinatal outcomes in Western women in the most

deprived neighbourhoods justifies additional attention in antenatal care. This approach could

include intensified prevention from existing health promoting programs in combination

with targeted social welfare. It is, however, difficult to extend this recommendation to non-

Western women, as our data and those of others do not show a straightforward relation

between social deprivation and perinatal outcome beyond the effect of being an ethnic

minority itself. Perhaps the ‘Western’ conventional deprivation indicators do not cover the

subtle ethnic specific pathways for adverse perinatal outcomes.

Non-Western women did not improve in perinatal outcome with increasing SI, but showed

an advantage over Western women in deprived neighbourhoods, possibly due to ‘protective’

knowledge systems. Future research could tap into these knowledge systems to evaluate

whether they can contribute to improvement in outcome with increasing SI in non-Western

women.

In addition, we also recommend more studies on this subject in European settings as most

studies on social deprivation and adverse perinatal outcomes have been conducted in

either the US or Canada which makes it unclear to what extent observations are valid for

European populations.

57

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

REFERENCES1. Mohangoo AD, Buitendijk SE, Hukkelhoven CW, et

al. Higher perinatal mortality in The Netherlands than in other European countries: the Peristat-II study. Ned Tijdschr Geneeskd. 2008;152:2718-27.

2. de Graaf JP, Ravelli AC, Wildschut HI, et al. Perinatal outcomes in the four largest cities and in deprived neighbourhoods in The Netherlands. Ned Tijdschr Geneeskd. 2008;152:2734-40.

3. Agyemang C, Vrijkotte TG, Droomers M, van der Wal MF, Bonsel GJ, Stronks K. The effect of neighbourhood income and deprivation on pregnancy outcomes in Amsterdam, The Netherlands. J Epidemiol Community Health. 2009;63:755-60.

4. Goedhart G, van Eijsden M, van der Wal MF, Bonsel GJ. Ethnic differences in preterm birth and its subtypes: the effect of a cumulative risk profile. BJOG. 2008;115:710-9.

5. Goedhart G, van Eijsden M, van der Wal MF, Bonsel GJ. Ethnic differences in term birthweight: the role of constitutional and environmental factors. Paediatr Perinat Epidemiol. 2008;22:360-8.

6. Poeran J, Denktas S, Birnie E, Bonsel GJ, Steegers EA. Urban perinatal health inequalities. J Matern Fetal Neonatal Med. 2011;24:643-6.

7. Fang J, Madhavan S, Bosworth W, Alderman MH. Residential segregation and mortality in New York City. Soc Sci Med. 1998;47:469-76.

8. Halpern D, Nazroo J. The ethnic density effect: results from a national community survey of England and Wales. Int J Soc Psychiatry. 2000;46:34-46.

9. Jonkers M, Richters A, Zwart J, Ory F, van Roos-malen J. Severe maternal morbidity among im-migrant women in the Netherlands: patients’ per-spectives. Reprod Health Matters. 2011;19:144-53.

10. Auger N, Giraud J, Daniel M. The joint influence of area income, income inequality, and immigrant density on adverse birth outcomes: a population-based study. BMC Public Health. 2009;9:237.

11. Elo IT, Culhane JF, Kohler IV, et al. Neighbourhood deprivation and small-for-gestational-age term births in the United States. Paediatr Perinat Epidemiol. 2009;23:87-96.

12. Farley TA, Mason K, Rice J, Habel JD, Scribner R, Cohen DA. The relationship between the neighbourhood environment and adverse birth outcomes. Paediatr Perinat Epidemiol. 2006;20:188-200.

13. O’Campo P, Burke JG, Culhane J, et al. Neighbor-hood deprivation and preterm birth among non-Hispanic Black and White women in eight geographic areas in the United States. Am J Epidemiol. 2008;167:155-63.

14. Timmermans S, Bonsel GJ, Steegers-Theunissen RP, et al. Individual accumulation of heterogeneous risks explains perinatal inequalities within deprived neighbourhoods. Eur J Epidemiol. 2011;26:165-80.

15. Genereux M, Auger N, Goneau M, Daniel M. Neighbourhood socioeconomic status, maternal education and adverse birth outcomes among mothers living near highways. J Epidemiol Community Health. 2008;62:695-700.

16. Janevic T, Stein CR, Savitz DA, Kaufman JS, Mason SM, Herring AH. Neighborhood deprivation and adverse birth outcomes among diverse ethnic groups. Ann Epidemiol. 2010;20:445-51.

17. Carstairs V, Morris R. Deprivation and health in Scotland. Health Bull (Edinb). 1990;48:162-75.

18. Perinatal care in The Netherlands 2007 (in Dutch: Perinatale zorg in Nederland 2007). Utrecht: The Netherlands Perinatal Registry; 2009. (online: http://www.perinatreg.nl/uploads/150/116/Jaarboek_Perinatale_Zorg_2007.pdf).

19. Kloosterman GJ. On intrauterine growth. Int J Gynaecol Obstet. 1970;8:895-912.

20. Gray R, Bonellie SR, Chalmers J, Greer I, Jarvis S, Williams C. Social inequalities in preterm birth in Scotland 1980-2003: findings from an area-based measure of deprivation. BJOG. 2008;115:82-90.

21. Fang J, Madhavan S, Alderman MH. Low birth weight: race and maternal nativity--impact of community income. Pediatrics. 1999;103:E5.

22. Grady SC. Racial disparities in low birthweight and the contribution of residential segregation: a multilevel analysis. Soc Sci Med. 2006;63:3013-29.

23. Neeleman J, Wessely S. Ethnic minority suicide: a small area geographical study in south London. Psychol Med. 1999;29:429-36.

58

SOCIA

L DEPRIVATIO

N A

ND

AD

VERSE PERINATA

L OU

TCOM

ES

4

24. Rabkin JG. Ethnic density and psychiatric hos-pitalization: hazards of minority status. Am J Psychiatry. 1979;136:1562-6.

25. Andres RL, Day MC. Perinatal complications associated with maternal tobacco use. Semin Neonatol. 2000;5:231-41.

26. Barnett R, Moon G, Kearns R. Social inequality and ethnic differences in smoking in New Zealand. Soc Sci Med. 2004;59:129-43.

27. Denktas S, Bonsel GJ, Van der Weg EJ, et al. An urban perinatal health programme of strategies to improve perinatal health. Matern Child Health J. 2011;16:1553-8.

28. Burris HH, Collins JW, Jr. Race and preterm birth--the case for epigenetic inquiry. Ethn Dis. 2010;20:296-9.

29. Kramer MR, Hogue CR. What causes racial disparities in very preterm birth? A biosocial perspective. Epidemiol Rev. 2009;31:84-98.

30. Collins JW, Jr., David RJ, Rankin KM, Desireddi JR. Transgenerational effect of neighborhood poverty on low birth weight among African Americans in Cook County, I llinois. Am J Epidemiol. 2009;169:712-7.

31. Crowder K, South SJ. Spatial Dynamics of White Flight: The Effects of Local and Extralocal Racial Conditions on Neighborhood out-Migration. Am Sociol Rev. 2008;73:792-812.

32. Schaake K, Burgers J, Mulder CH. Ethnicity at the Individual and Neighborhood Level as an Explanation for Moving Out of the Neighborhood. Popul Res Policy Rev. 2010;29:593-608.

33. Messer LC, Laraia BA, Kaufman JS, et al. The development of a standardized neighborhood deprivation index. J Urban Health. 2006;83:1041-62.

34. Moutquin JM. Classification and heterogeneity of preterm birth. BJOG. 2003;110 Suppl 20:30-3.

35. Aveyard P, Cheng KK, Manaseki S, Gardosi J. The risk of preterm delivery in women from different ethnic groups. BJOG. 2002;109:894-9.

36. Twisk JWR. Applied Multilevel Analysis. 4 ed. Cambridge: Cambridge University Press; 2010.

37. Rajaratnam JK, Burke JG, O’Campo P. Maternal and child health and neighborhood context: the selection and construction of area-level variables. Health Place. 2006;12:547-56.

38. Allebeck P. The prevention paradox or the inequal-ity paradox? Eur J Public Health. 2008;18:215.

PART II

Perinatal health in The Netherlands

60

5

Manuscript submitted for publication.

Population attributable risks of patient, child and organisational

risk factors for perinatal mortality

J. Poeran, G.J.J.M. Borsboom, J.P. De Graaf, E. Birnie, E.A.P. Steegers, G.J. Bonsel

62

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

ABSTRACTObjective Estimate the contributing role of maternal, child, and organisational risk factors

in perinatal mortality by calculating their population attributable risks (PAR).

Methods The primary dataset comprised all (n=1,020,749) singleton hospital births from

≥22 weeks’ gestation (The Netherlands Perinatal Registry 2000-2008). PARs for single and

grouped risk factors were estimated in four stages: (1) creating a duplicate dataset for each

PAR analysis in which risk factors of interest were set to the most favourable value (e.g.,

all women assigned ‘Western’ for PAR calculation of ethnicity); (2) in the primary dataset

an elaborate multilevel logistic regression model was fitted from which (3) the obtained

coefficients were used to predict perinatal mortality in each duplicate dataset; (4) PARs were

then estimated as the proportional change of predicted- compared to observed perinatal

mortality. Additionally, PARs for grouped risk factors were estimated by using incremental

values in two orders: after PAR estimation of grouped maternal risk factors, the resulting

PARs for grouped child, and grouped organisational factors were estimated, and vice versa.

Results The combined PAR of maternal, child and organisational factors is 94.4%, i.e., when

all factors are set to the most favourable value perinatal mortality is expected to be reduced

with 94.4%. Depending on the order of analysis, the PAR of maternal risk factors varies from

1.4-13.1%, and for child- and organisational factors 58.7-74.0% and 7.3-34.3%, respectively.

Conclusion The PAR of maternal, child and organisational factors combined is 94.4%.

Optimisation of organisational factors may achieve a 34.3% decrease in perinatal mortality.

63

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

INTRODUCTIONIn The Netherlands perinatal mortality rates exceed the European average.1,2 For long,

population factors such as the high maternal age at first childbirth, the high number of

multiple pregnancies, and the increasing proportion of non-Western pregnant women were

held responsible.1,2 These explanations were challenged as perinatal mortality remains high

after exclusion of these risk groups.3 Recent studies have thus addressed the contribution

of non-patient factors like the system of Dutch obstetric care with independently practicing

community midwives4, travel time to a hospital5, organisational characteristics of hospitals6,

and geographic factors, in particular urban deprivation7-9. However, of these studies neither

established the relative importance of the patient and non-patient risk factors, nor the

magnitude of the effect when unfavourable patient and non-patient risks coincide. A valid

estimate of attribution of these is essential to prioritise among available policy options

to reduce perinatal mortality. Our study aims to disentangle the contribution of patient

and non-patient risk factors to perinatal mortality, applying the population attributable

risk (PAR) concept to a comprehensive (2000-2008) national perinatal dataset. The PAR

concept is attractive as it generally quantifies the separate impact of a risk factor in

terms of proportion of total mortality (or other relevant outcome) accounted for by that

factor. The standard PAR approach assumes risk factors to act independently, and applies

straightforward formulae.10 The concept is considerably more complex if risk factors interact,

in a non-additive, conditional or time-dependent fashion. For convenience such interaction

effects are often ignored, or risk groups are excluded11, at the cost of conflicting or invalid

estimations of the impact of risk factors (in the perinatal context).

Our approach particularly accounts for interactions between the organisational features

of care provision and the fetal-maternal risk level. E.g., we account for the effect that travel

time to hospital5 for women referred during parturition may strongly depend on fetal

morbidity and maternal features: under perfect fetal-maternal conditions, travel time effects

may be minimal, while the effect may be sizable if the rate of unexpected transfers is high.

Likewise, we account for interaction between staffing level during out-of-office hours and

fetal-maternal characteristics.

We introduce a four-stage computational framework to estimate the relative importance

of interacting patient- and non-patient risk factors for perinatal mortality.

64

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

METHODS

Patient data

In this retrospective cohort study we derived data on maternal factors, child factors

and outcomes from The Netherlands Perinatal Registry. This registry contains complete

population-based information of >97% of all pregnancies in the Netherlands.12 On behalf

of the respective professional societies, source data are routinely collected by 94% of

midwives, 99% of gynaecologists and 68% of paediatricians including 100% of Neonatal

Intensive Care Unit paediatricians.12 (See website for detailed description: www.perinatreg.

nl). The registry contains 1,620,126 birth records for the period of 2000-2008.

We excluded multiple pregnancies (n=35,326), births with unknown gestational age

(n=17,768), births with unknown or erroneous zip code (n=20,804), and births supervised

exclusively by primary care midwives (n=525,479) as we assume hospital organisational

features to be of little or no importance in births under the exclusive supervision of primary

care midwives. Intra-uterine deaths (stillbirths, n=7,661) were excluded as their delivery

differs from normal cases.13

The final patient database consisted of 1,013,088 records.

Geographical data

Travel time to hospital was calculated as the travel time between a pregnant woman’s 4-digit

zip code and the hospital’s zip code; zip code covers an average of 4,000 inhabitants. Travel

time data were derived from a commercially available geographic information system

package acquired from the Geodan company (www.geodan.nl). We assume travel time to

hospital only to have an impact on women with unplanned births who were transferred

from home to hospital during parturition. Thus, hospital travel time was set to zero for

women with planned births, i.e., induction of labor (n=467,071) or a planned (primary)

caesarean section (n=98,135).

Hospital data

Hospital organisational data were collected separately from this study by one of the

investigators (JPDG), with the full support of the Dutch Society of Obstetricians and

Gynaecologists using a standardised (structured) interview for all (n=99) maternity units.

Response was 100%.6 Collected data consisted of 24 variables (table 5.1) with inevitable

65

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

Table 5.1 Overview of hospitals’ organisational data

Organisational feature Remarks

General  teaching hospital categorised into yes / no

Gynaecology departmentduration of daytime shifts categorised into durations of 10-12 hours and 7-9

hoursduration of evening / night shifts and the highest level of the professional who is attending these shifts

categorised into durations of ≤14 hours and ≥15 hours with professional levels categorised into two groups: (1) gynaecologist / gynaecologist in training and (2) non-in-training physician / midwife / nurse

the highest level of the professional who is attending the evening and night shifts during the week

categorised into five groups: (1) gynaecologist, (2) in-training and non-in-training resident gynaecology, (3) general emergency physician, (4) midwife and (5) nurse

the highest level of the professional who is attending the daytime shifts during weekends

categorised into five groups: (1) gynaecologist, (2) in-training and non-in-training resident gynaecology, (3) general emergency physician, (4) midwife and (5) nurse

the highest level of the professional who is attending the evening and night shifts during weekends

categorised into five groups: (1) gynaecologist, (2) in-training and non-in-training resident gynaecology, (3) general emergency physician, (4) midwife and (5) nurse

permitted to sleep during attending shifts categorised into yes / nogynaecologist on backup call during shifts categorised into yes / noprofessional doing rounds during weekends categorised into four groups: (1) gynaecologist,

(2) in-training and non-in-training resident gynaecology, (3) general emergency physician, (4) midwife and (5) nurse

in-hospital presence of the emergency operating theatre team

categorised into three groups: in-hospital presence during (1) 24 hours, (2) presence during daytime and evening and on-call during the night, (3) presence during daytime and on-call during the evening and night

in-hospital presence of an anesthaesiologist categorised into three groups: in-hospital presence during (1) 24 hours, (2) presence during daytime and evening and on-call during the night, (3) presence during daytime and on-call during the evening and night

the total number of obstetric caregivers (specialists, physicians, midwives, nurses)

a continuous number

annual number of deliveries continuous as well as grouped into three categories: (1) <1000, (2) 1000-2000, (3) >2000

combination of the annual number of deliveries with the gynaecologist:other obstetric caregiver ratio

categorised into six groups: (1) <1000 and ≤1:4, (2) <1000 and 1:4-1:6, (3) <1000 and >1:6, (4) 1000-2000 and ≤1:4, (5) 1000-2000 >1:4, (6) >2000 and all ratios

minimum duration of attending shift during weekends

a continuous number (range 12-72 hours)

Table 5.1 continues on next page.

66

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

Table 5.1 Continued

Organizational feature Remarks

average travel time in minutes from home to hospital for an on-call gynaecologist

a continuous number

maximum travel time in minutes from home to hospital for an on-call gynaecologist

a continuous number

number of caregivers present during attending shifts on weekday evenings and nights

a continuous number

number of caregivers present during attending shifts on weekends during the day

a continuous number

number of caregivers present during attending shifts on weekends during the evening and the night

a continuous number

Paediatrics departmentlevel of care for newborns categorised into four groups: (1) Neonatal Intensive

Care Unit (NICU), (2) post-NICU or incubator from 30 weeks of gestational age, (3) incubator from 32 weeks of gestational age, (4) initially first aid for neonate with suboptimal start with subsequent referral to a hospital with a higher level of care

the highest level of the professional at the paediatrics department who is attending the shifts outside office hours

categorised into six groups: (1) fellow neonatologist, (2) paediatrician, (3) in-training paediatrician, (4) non-in-training resident paediatrics, (5) general emergency physician, (6) no professional present but on-call

the lowest level of the professional at the paediatrics department who is attending the shifts outside office hours

categorised into six groups: (1) fellow neonatologist, (2) paediatrician, (3) in-training paediatrician, (4) non-in-training resident paediatrics, (5) general emergency physician, (6) no professional present but on-call

professional on-call at the paediatrics department

categorised into four groups: (1) no one on-call, (2) paediatrician on-call, (3) fellow neonatologist on-call, (4) other professional on-call

partial overlap. For that reason we used principal component analysis (PCA) as an accepted

technique to reduce the high number of interrelated explanatory variables to a limited

number of so-called principal components. PCA summarises the net information content

of overlapping data into a small number of independent, non-overlapping constructed

variables. Such a constructed variable consists of a weighted sum of the scores of the

observed variables, where each weight reflects the added informative value of each

observed variable to the constructed variable. The background and the technique itself

are described in more depth elsewhere.14 In the best case, 50% or more of all variability in

the original variables is covered by a limited number of (usually two or three) constructed

variables. The removal of overlap increases the power of the constructed variables as

67

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

determinants in explanatory regression analysis. Note that the PCA technique is descriptive

only: it can not be used for predictive analysis.

In our analysis the 24 organisation variables were summarised into two principal

components, factor 1 and factor 2, with a computed numerical score for each hospital.

Factor 1 can be interpreted as ‘scale size of the hospital’ as it primarily combines original

variables pointing to hospital size and the number of obstetric caregivers: the lower the

factor value, the larger the hospital and the higher the number of obstetric caregivers.

Factor 2 can be interpreted as ‘24-hour equality of service level’. This factor combines original

variables pointing to around-the-clock availability of qualified professionals from various

backgrounds. The two factors together cover about 70% of the information (variability)

contained in the original 24 variables.

We additionally computed a ‘hospital-specific intervention policy’ variable defined as the

hospital-specific percentage of primary caesarean sections for term breech presentation.

A low ‘primary-caesarean-section- for-term-breech’ percentage represents hospitals with

a more expectative approach in obstetric policy whereas a high percentage stands for

hospitals with a more proactive obstetric policy.

Outcome measures and risk factors

The main outcome was intrapartum or early neonatal death within 7 days after live birth,

for convenience further referred to as perinatal mortality.6 Risk factors were grouped into

maternal and child factors (patient factors) and organisational factors (non-patient factors).

Maternal factors were ethnicity (Western / non-Western of African descent / other non-

Western), parity (0 / 1-2 / >2) and maternal age (<25 / 25-29 / 30-34 / 35-39 / ≥40 years).

Child factors were gestational age (thirteen categories)6, congenital anomalies (yes / no),

small for gestational age (SGA: birthweight >10th / 2.3-10th / <2.3rd percentile)15,16 and fetal

presentation (cephalic / breech / transverse or other / unknown). Congenital anomalies are

recorded postpartum and classified through a standard coding system by organ system (8

categories, 71 subcategories).17

Travel time to the hospital was expressed in minutes (continuous). Other organisational

factors were: day and time of delivery (Saturday / Sunday / weekday, each subdivided

into three time slots 00:00-07:59 / 08:00-17:59 / 18:00-00:00), emergency referral during

parturition (yes / no), the two principal component factors, and the ‘primary-caesarean-

section-for-term-breech’ percentage. Emergency referral during parturition was defined

according to the caregiver providing delivery information in the registry.

68

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

Population attributable risk

The aim of this study is to estimate the population attributable risk (PAR) of maternal, child

and organisational risk factors for perinatal mortality. The PAR of a risk factor is the amount

of perinatal mortality that can be attributed to that particular risk factor among individuals

with the risk factor compared to those without.10 The standard PAR calculation is as follows:

PAR%=[(P*(RR−1))/(1+(P*(RR−1)))]*100. From this formula PAR estimations are expressed

as ‘the percentage or proportion of outcome accounted for’ by the risk factor involved. RR

stands for relative risk, and P for the prevalence of the risk factor in the studied population.

As said, PAR estimates from the above formula are subject to limitations if risk factors or

their effects interact. First, the formula is not additive if multiple risk factors act depend-

ently: a simple one-factor-at-a-time computation then leads to an unrealistic sum of PARs

being >100%. Second, computational complexity increases rapidly with more independent

factors, e.g., requiring simulation studies. Confounding adds to the complexity.

Logistic regression analysis enables to address several risk factors at a time, and indeed

has been used in some PAR studies.18 The resulting adjusted odds ratios (OR) are, however,

difficult to interpret, as the reference risk is unclear, and as this approach has no solution

for conditionally dependent factors.18 The method we present in this study addresses the

interaction problem. It only requires the researcher to be explicit on the presumed causal

priority of the related explanatory variables under study.11

Estimation of PARs

The PAR for single risk factors and groups of risk factors were estimated with a four-stage

approach (figure 5.1). In stage one, multiple duplicate datasets were created. In the duplicate

dataset the risk factor(s) of interest were eliminated by setting their values uniformly to

the most favourable value in terms of outcome. E.g., in the duplicate dataset used for the

PAR analysis of the single risk factor ‘ethnicity’, all women were assigned ‘Western’. The most

favourable value (categorical or continuous) was obvious in most variables. E.g., in case of

travel time to hospital, this value was zero; for the principal component factors 1 and 2, we

used the average value of the academic hospitals as optimum. For the primary-caesarean-

section-for-term-breech’ policy we replaced the observed percentage of primary caesarean

sections in term breech in a specific hospital by the national average, in those hospitals

where the observed percentage was lower than this average. Thus, a specific duplicate

dataset underlies each specific PAR estimate.

69

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

In stage two, a multilevel multivariable logistic regression model with perinatal mortality

as dependent variable was fitted using the original (unchanged) dataset. All determinants

mentioned above (=independent variables) were forced into the model, including

predefined interactions. Because of the multilevel structure of the data (individuals are

grouped per hospital) we used a multilevel logistic regression model which adjusts for the

possible clustering of particular individuals within hospitals (random intercept).

In stage three, the estimated beta-coefficients obtained in the multilevel model with the

original dataset, were used to predict -on the individual level- the probability of perinatal

mortality (‘PRED’ in figure 5.1) once a specific factor was eliminated. In each duplicate dataset

Figure 5.1 Four-stage procedural scheme for the estimation of PARs.

SUMMATION ofobserved cases of perinatal mortality

(‘OBS’)

SUMMATION ofpredicted individual perinatal mortality probabilities (‘PRED’)

PAR CALCULATION(in %) for eliminated risk factors

OBS - PREDOBS

X 100%

ORIGINAL DATASET

DUPLICATE DATASET

MULTIPLE DUPLICATEdatasets with risk factors ELIMINATED by setting the risk factors of interest to the most favourable value(s)

MULTILEVELREGRESSION MODEL

(forced inclusion of maternal, child and organisational factors

PREDICTING PERINATAL MORTALITY with defined risk factors eliminated

12 3

4

70

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

the role of one or more risk factors was eliminated by setting the value to the optimum. The

individual perinatal mortality probabilities were then summed to obtain a total predicted

perinatal mortality for each duplicate dataset with one or more risk factors eliminated. For

example, the predicted number of cases of perinatal mortality in a hypothetical world with

Western mothers only, all other factors equal, was estimated by applying the parameter

estimates from the multilevel model to the duplicate dataset in which all women were

assigned ‘Western’.

In stage four, PARs were calculated. The total predicted (by definition: lower) perinatal mortality

of each duplicate dataset was compared to the observed (‘OBS’ in figure 5.1) total mortality of

the original dataset. The PAR was then conventionally calculated as the proportional change

of the perinatal mortality (in %): e.g., if predicted mortality was 78% of the observed mortality,

the PAR was 22%. We did so for all single risk factors and groups of risk factors (maternal, child,

and organisational factors). If sets of risk factors were defined cumulative, this allowed for an

incremental analysis. E.g., if we start with estimating the PAR of ethnicity, followed by the PAR

of ethnicity and maternal age combined, the difference between these two PARs provides a

PAR of maternal age conditional on a population with only Western women.

The magnitude of the contribution of risk factors to perinatal mortality as computed by

this four-stage approach is presented in three ways:

1. the ‘univariable contribution’, where only the risk factor(s) at stake are set to the most

favourable values, leaving all other variables to their original level;

2. a ‘descending order contribution’, where we used a cumulative approach starting with

maternal, followed by child, and organisational factors, respectively. Note that in this

case the organisational PAR is estimated, assuming both maternal and child factors to

be optimal. For this method of calculation, the PAR of maternal factors is the same as

the ‘univariable contribution’ PAR.

3. ‘ascending order’ contribution’ with the reversed order of appearance of risk factors:

first organisational, then child, and finally maternal factors.

It can be expected that starting with organisational factors (‘ascending order’), leaving

the original level of mother and child risks unaltered, provides an equal or higher PAR,

compared to estimating the PAR of organisational factors conditional on a hypothetically

perfect mother-child population (‘descending order’). The comparison of results from (2) the

descending vs. (3) ascending order may thus be thought of as a policy choice to start with

either improvement of maternal and child factors vs. improvement of organisational factors.

71

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

SAS version 9.2 (SAS Institute Inc., Cary, NC) was used, with the GLIMMIX procedure to run

the random intercept multilevel models which were also used to calculate the predicted

values of perinatal mortality. Syntax can be obtained from the authors.

RESULTSIn our study population (table 5.2) most women are of Western origin (83.3%), primiparous

(52.9%) and 30-34 years old (38.3%). Preterm birth (<37 weeks’ gestation) was seen in 8.6%,

congenital anomaly in 3.0%, SGA in 10.6%, and a non-cephalic presentation was seen in 8.1%

of cases. Almost 15% of women had to travel 15 minutes or more to the hospital where they

gave birth, most children (42.4%) were born on a weekday between 08:00 and 17:59, 32.4%

of women were referred during parturition, and 12.5% of women gave birth in a hospital

with a ‘primary-caesarean-section- for-term-breech’ percentage of 75% or more. All academic

hospitals showed principal component factor values in the ‘best’ quartile (data not shown).

Table 5.2 Characteristics of the study population

  N Percentage

Study population 1,013,088 100,0%

Maternal factors

EthnicityWestern 844,340 83.3%Non-Western African descent 27,617 2.7%Non-Western other 141,131 13.9%

ParityPrimiparous (P0) 535,641 52.9%Multiparous (P1-P2) 425,288 42.0%Multiparous (P>2) 52,159 5.1%

Maternal age< 25 years 122,769 12.1%25-29 years 290,524 28.7%30-34 years 387,942 38.3%35-39 years 181,738 17.9%≥ 40 years 30,115 3.0%

Child factors

Gestational age22-27.6 weeks 4,876 0.5%28-31.6 weeks 8,417 0.8%32 weeks 4,238 0.4%

Table 5.2 continues on next page.

72

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

Table 5.2 Continued

  N Percentage

33 weeks 6,733 0.7%34 weeks 11,157 1.1%35 weeks 18,119 1.8%36 weeks 33,707 3.3%37 weeks 63,848 6.3%38 weeks 155,400 15.3%39 weeks 212,947 21.0%40 weeks 242,565 23.9%41 weeks 176,094 17.4%≥ 42 weeks 74,987 7.4%

Congenital anomaliesNo 982,273 97.0%Yes 30,815 3.0%

Small for gestational age (SGA)No SGA 905,919 89.4%Birthweight P2.3-P10 80,697 8.0%Birthweight < P2.3 26,472 2.6%

Fetal presentationCephalic 930,785 91.9%Breech 72,313 7.1%Transverse or other 9,036 0.9%Unknown 954 0.1%

Organisational factors

Travel time to hospital< 15 minutes 865,105 85.4%≥ 15 minutes 147,983 14.6%

Day and time of deliverySaturday 00:00-07:59 33,625 3.3%Saturday 08:00-17:59 55,509 5.5%Saturday 18:00-23:59 28,743 2.8%Sunday 00:00-07:59 33,750 3.3%Sunday 08:00-17:59 54,880 5.4%Sunday 18:00-23:59 28,434 2.8%Weekdays 00:00-07:59 173,463 17.1%Weekdays 08:00-17:59 429,398 42.4%Weekdays 18:00-23:59 175,286 17.3%

Referral during parturitionNo 684,495 67.6%Yes 328,593 32.4%

Hospital ‘elective-caesarean-section- for-term-breech’ percentage< 75% 886,893 87.5%≥ 75% 126,195 12.5%

73

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

Multivariable multilevel logistic regression model using the original dataset

Effect estimates from the multivariable multilevel logistic regression model are shown

in table 5.3. For maternal factors, only parity is significantly associated with perinatal

mortality, with the highest risk for multiparous (P>2) women (OR 1.57; CI 1.38-1.78). All

child factors are significantly associated with perinatal mortality. Gestational ages <37

weeks have the highest risks for perinatal mortality (OR range 3.71 to >1,000); the highest

risk for term pregnancies is seen for 37 weeks of gestation (OR 2.35; CI 2.01-2.75). Infants

with congenital anomalies have an 11 times increased risk (OR 11.05; CI 10.25-11.91),

and a birthweight < P2.3 has an almost 7 times increased risk (OR 6.79; CI 6.10-7.56) for

perinatal mortality. Increased risks are also observed for non-cephalic fetal presentations

(OR range 1.53-2.20). For organisational factors only the principal component factors are

Table 5.3 Multivariable multilevel logistic regression model

  aOR CI  P-value

Maternal factors

Ethnicity 0.083Western [ref] 1.00Non-Western African descent 1.10 0.94 1.28Non-Western other 1.10 1.01 1.20

ParityPrimiparous (P0) [ref] 1.00 <.0001Multiparous (P1-P2) 1.31 1.23 1.40Multiparous (P>2) 1.57 1.38 1.78

Maternal age 0.085< 25 years 1.05 0.95 1.1625-29 years [ref] 1.0030-34 years 1.07 0.99 1.1635-39 years 1.13 1.03 1.24≥ 40 years 0.96 0.81 1.15

Child factors

Gestational age <.000122-27.6 weeks >1,000 >1,000 >1,00028-31.6 weeks 43.38 37.28 50.4832 weeks 16.68 13.52 20.5933 weeks 12.42 10.22 15.0934 weeks 7.35 6.07 8.91

Table 5.3 continues on next page.

74

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

Table 5.3 Continued

  aOR CI  P-value

35 weeks 5.54 4.64 6.6236 weeks 3.71 3.14 4.3737 weeks 2.35 2.01 2.7538 weeks 1.36 1.18 1.5739 weeks 1.05 0.91 1.2140 weeks [ref] 1.0041 weeks 1.34 1.16 1.54≥ 42 weeks 1.35 1.12 1.63

Congenital anomalies <.0001No [ref] 1.00Yes 11.05 10.25 11.91

Small for gestational age (SGA)No SGA [ref] 1.00 <.0001Birthweight P2.3-P10 2.41 2.18 2.65Birthweight < P2.3 6.79 6.10 7.56

Fetal presentation <.0001Cephalic [ref] 1.00Breech 1.53 1.41 1.66Transverse or other 1.54 1.29 1.83Unknown 2.20 1.26 3.86

Organisational factors

Travel time to hospital 1.00 0.99 1.00 <.0001

Day and time of delivery <.0001Saturday 00:00-07:59 1.49 1.27 1.75Saturday 08:00-17:59 1.11 0.97 1.29Saturday 18:00-23:59 1.34 1.12 1.60Sunday 00:00-07:59 1.40 1.19 1.64Sunday 08:00-17:59 1.31 1.14 1.50Sunday 18:00-23:59 1.16 0.96 1.39Weekdays 00:00-07:59 1.42 1.31 1.55Weekdays 08:00-17:59 [ref] 1.00Weekdays 18:00-23:59 1.29 1.19 1.41

Referral during parturition <.0001No [ref] 1.00Yes 1.24 1.14 1.34

Hospital ‘elective-caesarean-section-for-term-breech’ percentage

0.99 0.98 0.99 <.0001

Hospital organisational featuresPrincipal component (factor 1) 1.01 0.91 1.11 0.915Principal component (factor 2) 0.98 0.89 1.09 0.752Principal component (factor 1) ^2 0.86 0.77 0.97 0.012Principal component (factor 2) ^2 0.99 0.95 1.04 0.746

75

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

not significantly associated with perinatal mortality (OR 1.01; CI 0.91-1.11 and OR 0.98; CI

0.89-1.09 for factor 1 and 2, respectively). Births on Saturday nights and weekday nights

have a >40% increased risk for perinatal mortality (OR 1.49; CI 1.27-1.75 and OR 1.42; CI

1.31-1.55, respectively), referral during parturition has an almost 25% increased risk (OR 1.24;

CI 1.14-1.34). The hospital ‘primary-caesarean-section- for-term-breech’ percentage is also

significantly associated with perinatal mortality (0.99; CI 0.98-0.99). For every percentage

increase in ‘primary-caesarean-section- for-term-breech’, perinatal mortality decreases with

1%. Thus, a hospital with a higher percentage of primary caesarean sections on average

has a lower rate of perinatal mortality adjusted for other factors.

PAR results

Table 5.4 lists PARs of single risk factors (upper part) and groups of risk factors (‘univariable

contribution’; lower part). Of all maternal factors, parity has the highest univariable PAR

(8.9%). In other words: if all women in our dataset would be primiparous perinatal mortality

would be reduced by 8.9%, all other things equal. Of all child factors, gestational age (72.2%)

and congenital anomalies (22.7%) have the highest univariable PAR. Hospital organisational

factors captured in the ‘primary-caesarean-section- for-term-breech’ percentage and

the principal component factors have the highest univariable PAR (13.2% and 16.4%,

respectively) of the organisational factors. If every hospital is set to the between-hospital

average ‘primary-caesarean-sections-for-term-breech’ percentage, perinatal mortality would

decline with 13.2%. If all hospitals are set to the most favourable principal component

Table 5.4 Observed and predicted perinatal mortality with subsequent PAR estimations of single risk factors (upper part) and groups of risk factors (lower part)

  N OBS-PRED PAR (%)

Observed cases of perinatal mortality 6,269 - 100,0%

Predicted cases of perinatal mortality

Maternal factors

EthnicityEveryone Western 6,184 85 1.4%

ParityEveryone nulliparous 5,711 558 8.9%

Maternal ageEveryone 25-29 years 6,047 222 3.5%

Table 5.4 continues on next page.

76

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

Table 5.4 Continued

  N OBS-PRED PAR (%)

Child factors

Gestational ageEveryone 40 weeks 1,740 4,529 72.2%

Congenital anomaliesNo congenital anomalies 4,848 1,421 22.7%

Small for gestational age (SGA)No SGA 5,554 715 11.4%

Fetal presentationEveryone cephalic 5,881 388 6.2%

Organisational factors

Travel time to hospitalEveryone no travel time 6,458 -189* *

Day and time of deliveryEveryone on weekdays 08:00-17:59 5,564 705 11.2%

Referral during parturitionNo referrals 6,076 193 3.1%

Hospital ‘elective-caesarean-section- for-term-breech’ percentageEvery hospital the average % 5,444 825 13.2%

Principal component factorsAll principal component factors set to the most favorable value 5,240 1,029 16.4%

Groups of factors

All maternal factors set to the most favorable value 5,450 819 13.1%

All child factors set to the most favorable value 1,004 5,265 84.0%

All organisational factors set to the most favorable value 4,120 2,149 34.3%

All mother and child factors set to the most favorable value 811 5,458 87.1%

All child and organisational factors set to the most favorable value 438 5,831 93.0%

All maternal, child and organisational factors set to the most favorable value

352 5,917 94.4%

 * In table 5.3 ‘travel time to hospital’ has an OR of 0.995 representing a 0.5% decreased risk of perinatal mortality for every minute of extra travel time. Therefore, predicted mortality > observed mortality when everyone in the duplicate dataset is assigned a 0 for travel time. This would result in a negative PAR according to our calculation method.

factor value, perinatal mortality would be reduced with 16.4%. According to the ‘univariable

contribution’ estimation method, the estimated PAR of all maternal factors combined is

13.1%. Univariable contribution of the child factors combined and the organisational factors

combined is 84.0% and 34.3%, respectively. The PAR of all maternal, child and organisational

factors together is 94.4%. In other words: when all factors are set to the most favourable

77

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

value the perinatal mortality is expected to be reduced with 94.4%, pointing to a high

explanatory power of the complete set of variables in this analysis.

Table 5.5 compares results from the ‘descending order contribution’ vs. the ‘ascending order

contribution’. The former estimation method results in a PAR of 13.1%, 74.0% and 7.3% for

maternal, child and organisational factors, respectively. The latter estimation method results

in a PAR of 34.3%, 58.7% and 1.4% for organisational, child and maternal factors, respectively.

Hence, depending on the estimation method, the PAR of maternal factors varies from 1.4 to

13.1%; for child factors from 58.7 to 74.0%, and for organisational factors from 7.3 to 34.3%.

DISCUSSION

Principal fi ndings

Combining a multilevel logistic regression approach with a specific ‘what-if’ framework,

we were able to estimate the impact of maternal, child and organisational risk factors

on perinatal mortality. We estimated the combined population attributable risk (PAR) of

Table 5.5 PARs of groups of risk factors estimated by two methods: ‘descending order contribution’ (left) and ‘ascending order contribution’ (right)

 

  ↓  

MATERNAL FACTORS 13.1% 94.4% MATERNAL FACTORS

↓ +74.0% +1.4% ↑CHILD FACTORS 87.1% 93.0% CHILD FACTORS

↓ +7.3% +58.7% ↑ORGANISATIONAL FACTORS 94.4% 34.3% ORGANISATIONAL FACTORS

 

   ↑

78

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

patient- (maternal and child) and non-patient (organisational) factors to be 94.4%. Thus,

almost all perinatal mortality could be explained statistically. Gestational age showed the

highest single contribution (PAR of 72.2%), but also organisational factors as single factors

ranked high (PAR of 34.3%). In estimating the risk attributions for combined maternal, child

and organisational factors, results depended on the causal priority given on the various

risk groups. By selecting a ‘descending order’ the PARs of maternal, child and organisational

factors are 13.1%, 74.0% and 7.3%, respectively. The reverse order provides PARs of 34.3%,

58.7% and 1.4%, which can be translated into the following: optimising organisational

factors with the current pregnant population has a potential of a one third decrease in

perinatal mortality.

Also, our data showed the overriding importance of decreasing preterm births, and a

distinct >200% risk of a birth at 37.0-37.6 weeks of gestation (table 5.3, OR 2.35), considered

‘at term’6. This finding suggests reconsidering the threshold of ‘term’ pregnancy from 37

weeks to 38 weeks.19,20

Other PAR studies

Previous studies describing PARs for perinatal mortality have used multivariable regression

models with the resulting adjusted odds ratio (OR) replacing the relative risk (RR) in the

conventional PAR formula (depicted in the methods section).18,21-23 Two previous studies have

used data from The Netherlands Perinatal Registry to calculate PARs for perinatal mortality,

however with each a different approach.11,22 Ravelli and colleagues calculated adjusted PARs

for perinatal mortality (stillbirths included) through multivariable regression and adjusted

odds ratios (see above). Focusing on maternal factors only, they showed nulliparity to have

the highest PAR for perinatal mortality (14.8%), considerably more than e.g. maternal age

(PAR 1.4-3.6%), ethnicity (PAR 6.6%), and gender of the child (PAR 3.6%).22 The difference

with our PARs arises from the absence of child and organisational factors, the different

computational procedure (see Methods), and the inclusion of stillbirths. Gijsen et al. used

a method comparable to our multilevel and ‘observed-expected’ method in estimating

perinatal mortality attributable to the effect of an off-hours delivery.11 They found an overall

PAR of 12.5% of evening and night time deliveries for intrapartum and early neonatal

mortality. A limitation of this study is the exclusion of several groups at risk for intrapartum

and early neonatal mortality, e.g., small for gestational age infants and severe congenital

anomalies which most likely are more vulnerable to suboptimal care.

79

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

Strengths and weaknesses

An important strength of our study is the usage of a validated national database (The

Netherlands Perinatal Registry) with complete coverage of all deliveries over a long period

of time (2000-2008). Second, compared to previous studies, our explanatory model included

more detailed information on organisational factors from all Dutch hospitals, built on a

previous study on hospital related effects.6,11,22 Third, we used a novel method to calculate

PARs in a multivariable context, and demonstrated a strong effect of the researcher’s choice

of explanatory variables (‘ascending and ‘descending order contribution’) and on the order

of including these variables in computing PARs.

Provided that appropriate choices are made, we assume the resulting multivariable PARs

to be more accurate compared to PAR estimations from the conventional PAR-formula

which is unsuitable in case of interacting variables.18 Generally, a multivariable PAR provides

additional information beyond measures on the strength of association.4-6,8,9,24-26 For example:

a placental abruption may have a high OR for perinatal mortality; however, on the population

level its attributable risk is limited as the condition is very rare; PARs account for this.10

This study also has limitations. First, the lack of detailed data on maternal lifestyle factors

in The Netherlands Perinatal Registry, in particular smoking, may contribute to a too small

PAR of maternal factors.27 By including SGA and gestational age in our study model part

of this maternal smoking effect is switched to these child factors. Still, in view of the low

PAR of maternal factors, our study suggests that in the perinatal mortality context, casemix

adjustment for maternal factors is of less value than often assumed.1-3 A second limitation

refers to the modifiability of risk factors, which is not addressed by our method. In public

health, PARs are generally calculated for modifiable risk factors only as the concept is used

to judge priorities for intervention programs. Some risk factors in our study obviously are

non-modifiable (e.g., parity or ethnicity), and others only partially. Still, the small PARs

on parity and ethnicity may serve the discussion on relevance and prioritisation. Finally,

we by intent excluded births under supervision of independently practicing midwives

(n=525,479 from a total of 1,620,126). Our results only pertain to hospital births; as in most

Western countries births take place in hospital this selection may account for international

generalisability of our results.

Possible implications

In assuming organisational factors to be modifiable at least partially on the long term, our

study suggests a major decrease in perinatal mortality (PAR of 34%) after these factors have

80

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

been optimised. Possible policies include ensuring (e)quality of obstetric care during off-

hours, and a proactive approach in obstetric policy. Centralisation of obstetric care may be

deemed necessary in optimising organisational factors ensuring 24-hour high level obstetric,

anaesthesia, and optimal neonatal resuscitation coverage.28 Next to organisational factors,

the PAR estimations of child factors suggest substantial reduction of perinatal mortality on

the population level through measures decreasing preterm birth rates. Foremost, methods

to improve detection of (risks for) preterm are needed which may be achieved through

standardised antenatal risk scoring systems, or prediction algorithms for preterm birth.29,30

The threshold of term may be shifted to 38 weeks gestational age.

Conclusion

We quantified the overall impact on in-hospital perinatal mortality of maternal, child, and

organisational risk factors, adjusting for their mutual dependencies. The so-called population

attributable risk (PAR) of these three groups of factors combined is 94.4%, where gestational

age has the highest single impact. The PARs of risk factors separately vary by the method

of estimation, in particular the order of adjustment which in turn depends on the intended

use. Focusing on the role of hospital organisational factors, the PAR results suggest that

a change towards the theoretical organisation optimum with an otherwise unchanged

pregnant population provides a 34.3% decrease in in-hospital perinatal mortality.

REFERENCES1. Buitendijk SE, Nijhuis JG. High perinatal mortality

in the Netherlands compared to the rest of Europe. Ned Tijdschr Geneeskd. 2004;148:1855-1860.

2. Mohangoo AD, Buitendijk SE, Hukkelhoven CW, et al. Higher perinatal mortality in The Netherlands than in other European countries: the Peristat-II study. Ned Tijdschr Geneeskd. 2008;152:2718-2727.

3. Anthony S, Jacobusse GW, van der Pal-de Bruin KM, Buitendijk S, Zeitlin J, Factors E-PWGoR. Do differences in maternal age, parity and multiple births explain variations in fetal and neonatal mortality rates in Europe?--Results from the EURO-PERISTAT project. Paediatr Perinat Epide-miol. 2009;23:292-300.

4. Evers AC, Brouwers HA, Hukkelhoven CW, et al. Perinatal mortality and severe morbidity in low and high risk term pregnancies in the Netherlands: prospective cohort study. BMJ. 2010;341:c5639.

5. Ravelli AC, Jager KJ, de Groot MH, et al. Travel time from home to hospital and adverse perinatal outcomes in women at term in the Netherlands. BJOG. 2011;118:457-465.

6. de Graaf JP, Ravelli AC, Visser GH, et al. Increased adverse perinatal outcome of hospital delivery at night. BJOG. 2010;117:1098-1107.

7. Poeran J, Denktas S, Birnie E, Bonsel GJ, Steegers EA. Urban perinatal health inequalities. J Matern Fetal Neonatal Med. 2011;24:643-646.

81

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

8. Ravelli AC, Steegers EA, Rijninks-van Driel GC, et al. Differences in perinatal mortality between districts of Amsterdam. Ned Tijdschr Geneeskd. 2011;155:A3130.

9. Timmermans S, Bonsel GJ, Steegers-Theunissen RP, et al. Individual accumulation of heterogeneous risks explains perinatal inequalities within deprived neighbourhoods. Eur J Epidemiol. 2011; 26:165-180.

10. Altman DG. Strategies for community health intervention: promises, paradoxes, pitfalls. Psy-chosom Med. 1995;57:226-233.

11. Gijsen R, Hukkelhoven CW, Schipper CM, Ogbu UC, de Bruin-Kooistra M, Westert GP. Effects of hospital delivery during off-hours on perinatal outcome in several subgroups: a retrospective cohort study. BMC Pregnancy Childbirth. 2012;12:92.

12. Perinatal care in The Netherlands 2008 (in Dutch: Perinatale zorg in Nederland 2008). Utrecht: The Netherlands Perinatal Registry; 2011.

13. Stillbirth Collaborative Research Network Writing G. Causes of death among stillbirths. JAMA. 2011;306:2459-2468.

14. Burstyn I. Principal component analysis is a powerful instrument in occupational hygiene inquiries. Ann Occup Hyg. 2004;48:655-661.

15. Manning, FA. Intrauterine growth retardation. In: Manning, FA, ed. Fetal medicine: Principles and practice. Appleton and Lange, Norwalk, CT 1995.

16. Mikolajczyk RT, Zhang J, Betran AP, et al. A global reference for fetal-weight and birthweight percentiles. Lancet. 2011;377:1855-1861.

17. Rietberg CCT. Term breech delivery in The Netherlands; chapter 7. Thesis 2006, University of Utrecht. (online available: http://igitur-archive.library.uu.nl/dissertations/2006-1031-200813/c7.pdf ).

18. Benichou J. A review of adjusted estimators of attributable risk. Stat Methods Med Res. 2001; 10:195-216.

19. Fleischman AR, Oinuma M, Clark SL. Rethinking the definition of “term pregnancy”. Obstet Gynecol. 2010;116:136-139.

20. Reddy UM, Bettegowda VR, Dias T, Yamada-Kushnir T, Ko CW, Willinger M. Term pregnancy: a period of heterogeneous risk for infant mortality. Obstet Gynecol. 2011;117:1279-1287.

21. Hutcheon JA, McNamara H, Platt RW, Benjamin A, Kramer MS. Placental weight for gestational age and adverse perinatal outcomes. Obstet Gynecol.119:1251-1258.

22. Ravelli AC, Tromp M, van Huis M, et al. Decreasing perinatal mortality in The Netherlands, 2000-2006: a record linkage study. J Epidemiol Community Health. 2009;63:761-765.

23. Cruz-Anguiano V, Talavera JO, Vazquez L, et al. The importance of quality of care in perinatal mortality: a case-control study in Chiapas, Mexico. Arch Med Res. 2004;35:554-562.

24. de Graaf JP, Ravelli AC, Wildschut HI, et al. Perinatal outcomes in the four largest cities and in deprived neighbourhoods in The Netherlands. Ned Tijdschr Geneeskd. 2008;152:2734-2740.

25. Tromp M, Eskes M, Reitsma JB, et al. Regional perinatal mortality differences in the Netherlands; care is the question. BMC Public Health. 2009;9: 102.

26. van den Hooven EH, Pierik FH, de Kluizenaar Y, et al. Air pollution exposure during pregnancy, ultrasound measures of fetal growth, and adverse birth outcomes: a prospective cohort study. Environ Health Perspect. 2012;120:150-156.

27. Andres RL, Day MC. Perinatal complications associated with maternal tobacco use. Semin Neonatol. 2000;5:231-241.

28. Heller G, Richardson DK, Schnell R, Misselwitz B, Kunzel W, Schmidt S. Are we regionalized enough? Early-neonatal deaths in low-risk births by the size of delivery units in Hesse, Germany 1990-1999. Int J Epidemiol. 2002;31:1061-1068.

29. Davey MA, Watson L, Rayner JA, Rowlands S. Risk scoring systems for predicting preterm birth with the aim of reducing associated adverse outcomes. Cochrane Database Syst Rev. 2011:CD004902.

30. Schaaf JM, Ravelli AC, Mol BW, Abu-Hanna A. Development of a prognostic model for predict-ing spontaneous singleton preterm birth. Eur J Obstet Gynecol Reprod Biol. 2012;164:150-155.

82

POPU

LATION

ATTRIBUTA

BLE RISKS FOR PERIN

ATAL M

ORTA

LITY

5

6The ‘Healthy Pregnancy 4 All’ Study:

design and cohort profile

S. Denktaş, J. Poeran, L.C. de Jong-Potjer, S.F. van Voorst, A.A. Vos, A.J.M. Waelput, E. Birnie, G.J. Bonsel, E.A.P. Steegers

Manuscript submitted for publication.

84

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

ABSTRACTThe design of a national study designated as ‘Healthy Pregnancy 4 All’ (HP4All) is presented.

This study combines epidemiologic and health services research to evaluate the effectiveness

of two obstetric interventions: (1) programmatic preconception care (PCC) provision to

reduce the risks and risk load related adverse pregnancy outcome in a population-based

prospective cohort study; (2) a score card-based antenatal risk assessment in a cluster

randomised controlled trial for medical and non-medical risks related to adverse pregnancy

outcome, followed by patient-tailored multidisciplinary care pathways. Altogether, 14

municipalities/regions were selected to participate according to socio-demographic data

(high risk load), perinatal outcome data (high adverse outcome prevalence), and some

specific data for either PCC or antenatal risk assessment. In total, 839 women need to be

included in the preconception care study and 7,000 women need to be included in the

antenatal risk assessment study. Data are collected by physical examinations, questionnaires,

interviews, and biological samples. Preparations started in the spring of 2011. Participant

recruitment and complete data collection started in the fall of 2012.

85

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

INTRODUCTIONPerinatal mortality rates in the Netherlands are high and decline slowly compared to other

European countries.1-3 More risks and a higher risk load for adverse outcomes were found

for women living in socially deprived areas.4 With the support of the Ministry of Health

and Welfare a nationwide study, called ‘Healthy Pregnancy 4 All’ (HP4All), was developed

to provide evidence based strategies at an early stage to improve pregnancy outcome.

Several municipal pilot studies in the city of Rotterdam provided the framework for this

national study.5

Main objectives

The main HP4All study objective is to evaluate the effectiveness of two interventions and

their associated preventive strategies in either the preconception period or the antenatal

period to reduce adverse pregnancy outcome. Accordingly, two sub-studies are specified:

a population-based prospective cohort study focusing on the effectiveness of customised

preconception care (PCC) and a cluster randomised trial focusing on the early identification

of groups at risk for adverse pregnancy outcomes (in particular, preterm birth and small

for gestational age) and the related effectiveness of score-card based early antenatal risk

assessment with the so-called ‘R4U’ (Rotterdam Reproductive Risk Reduction) score card.

Rationale

The rationale of the PCC sub-study originates from increasing evidence showing the

critical influence of embryonic development and placentation during early pregnancy on

pregnancy outcome.6-8 Risks influencing this early pregnancy phase are optimally modified

in the preconception period.6,9,10 The Dutch Health Council recommended (2007) to integrate

general PCC in the health care system.11 The Minister of Health, however, advised to first

evaluate the utilisation and effectiveness of PCC for high risk groups in the HP4All study,

before the nationwide implementation of PCC in Dutch obstetric care can be considered.

The rationale of the second sub-study on early antenatal risk assessment originates from the

unique Dutch system of obstetric care which has three risk-based levels of care: primary care

for low risk pregnancies and deliveries, provided by independently practicing community

midwives, and secondary/tertiary care for high risk pregnancies provided by obstetricians.12

As the level of care depends on the distinction between assumed low risk and high risk

pregnancies, antenatal risk assessment by primary care community midwives is an important

86

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

part of Dutch obstetric care.12 Although social deprivation has been shown to contribute to

adverse perinatal health in the Netherlands, standard risk assessment does not include the

assessment of social risks of perinatal health.4,5,13,14 In addition, subsequent patient-tailored

social care pathways are lacking. Therefore, in the new antenatal risk assessment tool (‘R4U’)

explicitly social and medical risk factors are taken into account as part of the HP4All study.

Population-based cohort studies, e.g., the Generation R15 and ABCD16 studies have

contributed to our knowledge of various health problems in pregnancy and childhood

and their lasting impact on health in later life. Moreover, studies using a large national

Dutch database (The Netherlands Perinatal Registry) showed increased adverse pregnancy

outcome in large urban areas, in particular in deprived neighbourhoods.13,17 Also, four

specific morbidities appear to precede perinatal mortality in 85% of cases, the so-called ‘Big4’

morbidities.18,19 These are: congenital anomalies (list defined), preterm birth (<37th week

of gestation), small for gestational age (SGA, birth weight <10th percentile for gestational

age) or low Apgar score (<7, 5 minutes after birth). Taking advantage of knowledge from

these cohort studies, the HP4All study will study the effectiveness of newly introduced

evidence-based care strategies for early identification and customised care provision to

women at risk.

Below, we first describe the selection of geographical areas most suitable for the

interventions. Next, we describe the designs of the preconception care and the antenatal

risk assessment studies.

SELECTION OF THE PARTICIPATING MUNICIPALITIES/REGIONSThe first step in HP4All was the identification of the geographical unit in which the

aforementioned sub-studies would preferably be carried out. We used a national geographic

information system (GIS) to divide The Netherlands into 62 municipalities/regions, being

the 50 cities with > 70,000 inhabitants and the 12 provinces (excluding the 50 previously

selected cities). The second step involved the selection of municipalities/regions in which

to carry out the sub-studies, based on multiple criteria which are relevant to either the

preconception care sub-study, to the enhanced antenatal risk assessment sub-study,

or to both. The final step dealt with additional conditions and the final selection of the

participating municipalities/regions which is elaborated on below.

87

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Selection criteria

Initially, we selected municipalities/regions according to socio-demographic parameters

associated with high risk load (maternal age, parity, ethnicity, and socioeconomic status)

and perinatal outcome data (overall ‘Big4’ and perinatal mortality prevalence). Before the

municipalities/regions could be selected, specific parameters relevant to either sub-study

PCC or antenatal risk assessment were added.

For the PCC sub-study these criteria were (1) proportion of women having their first antenatal

booking visit at ≥14 weeks of gestational age, and prevalences of (2) congenital anomalies

and of (3) SGA. A timely first antenatal booking is important because the opportunities

for prenatal screening and interventions (i.e., lifestyle advice and changes) as well as the

effectiveness of those interventions are larger in an early fetal stage. Congenital anomaly and

SGA prevalences are considered to be indicative for a region’s periconceptional health status.

For the antenatal risk assessment sub-study, additional criteria were (1) overall perinatal

mortality rates, (2) perinatal mortality amongst women with ‘Big4’ pregnancies, and (3)

prevalence of SGA and prematurity.

Data sources

The division of The Netherlands into 62 municipalities/regions was based on 4-digit postal

codes areas. Data were provided by the Falk company (www.falk.nl), the National Public

Health Authority, and the Statistics Netherlands organisation (CBS, www.cbs.nl). Information

on socioeconomic status (SES, determined in 2006) per postal code area was obtained from

the Social and Cultural Planning Office (SCP, www.scp.nl). Data on pregnancy and perinatal

outcome were derived from The Netherlands Perinatal Registry (2000-2008), containing

information of more than 97% of all pregnancies in The Netherlands.20 These data are

routinely collected by 94% of midwives, 99% of gynaecologists and 68% of paediatricians

including 100% of Neonatal Intensive Care Unit paediatricians.20 Table 6.1 shows the

demographic characteristics of the so-called ‘G4-cities’, i.e., the four largest cities: Amsterdam,

Rotterdam, The Hague, Utrecht, and the rest of the Netherlands. Compared to the rest of The

Netherlands, the ‘G4’-cities have a larger proportion of non-Western women (43% vs. 11.3%),

more teenage pregnancies (2.8% vs. 1.5%), and more women in low SES neighbourhoods

(59.2% vs. 19.0%). Considerably more women live in deprived neighbourhoods (32.5% vs.

1.3%) and the overall adverse perinatal outcome is worse in ‘G4-cities’, as illustrated by a

‘Big4’ prevalence of 20.5% compared to 18.1%.

88

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Table 6.1 Demographic characteristics of the study population by yes/no ‘G4-cities’ (the four largest cities) with percentages in brackets

  

G4-cities Netherlands minus G4-cities

Total

No. of pregnancies during study period 245,445 (100.0) 1,338,420 (100.0) 1,583,865 (100.0)

ParityPrimiparous 121,592 (49.5) 607,953 (45.4) 729,545 (46.1)Multiparous 123,853 (50.5) 730,467 (54.6) 854,320 (53.9)

EthnicityWestern 139,786 (57.0) 1,186,772 (88.7) 1,326,558 (83.8)Non-Western 105,659 (43.0) 151,648 (11.3) 257,307 (16.2)

Maternal age< 20 years 6,987 (2.8) 19,861 (1.5) 26,848 (1.7)20-24 years 34,864 (14.2) 127,013 (9.5) 161,877 (10.2)25-29 years 61,354 (25.0) 395,138 (29.5) 456,492 (28.8)30-34 years 85,444 (34.8) 535,927 (40.0) 621,371 (39.2)≥ 35 years 56,796 (23.1) 260,481 (19.5) 317,277 (20.0)

Socioeconomic 'status score'<p20 145,367 (59.2) 254,607 (19.0) 399,974 (25.3)p20-p80 58,641 (23.9) 853,074 (63.7) 911,715 (57.6)>p80 41,437 (16.9) 230,739 (17.2) 272,176 (17.2)

NeighbourhoodNon-deprived 165,658 (67.5) 1,320,392 (98.7) 1,486,050 (93.8)Deprived 79,787 (32.5) 18,028 (1.3) 97,815 (6.2)

Perinatal outcomes**Congenital anomalies 5,233 (2.1) 33,159 (2.5) 38,392 (2.4)Preterm birth 15,673 (6.4) 81,646 (6.1) 97,319 (6.1)Small for gestational age 27,724 (11.3) 125,175 (9.4) 152,899 (9.7)Apgar score <7 (5 minutes after birth) 3,385 (1.4) 14,818 (1.1) 18,203 (1.1)

Any Big4** 50,267 (20.5) 242,697 (18.1) 292,964 (18.5)

Fetal mortality† 1,478 (0.6) 6,718 (0.5) 8,196 (0.5)

Intrapartum mortality 458 (0.2) 2,126 (0.2) 2,584 (0.2)

Neonatal mortality†† 761 (0.3) 3,547 (0.3) 4,308 (0.3)

Perinatal mortality‡ 2,697 (1.1) 12,391 (0.9) 15,088 (1.0)

** Individual 'Big4' morbidities do not add up to 'any Big4' as women can have >1 'Big4' morbidity.† From 22 weeks of gestational age.†† 0-7 days postpartum.‡ Total of fetal, intrapartum and neonatal mortality.

89

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Perinatal mortality and ‘Big4’ prevalence

Figures 6.1 and 6.2 illustrate the geographical distribution of perinatal mortality rates, and

the prevalence rate of ‘Big4’ (per 1,000) respectively. Various shades of red represent the

different prevalence classes, the darker the shade the more prevalent the adverse outcome.

The classes are based on the distribution of the rates: the middle three classes comprise 95%

(2 standard deviations) of the outcome levels; the middle class comprises 68%. Both figures

show large geographical inequalities in adverse perinatal outcomes on the national level.

Figure 6.1 Absolute prevalence of perinatal mortality per 1,000 births.

90

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Comparison municipalities/regions

We additionally compared these outcomes across regions after direct standardisation21 for

population differences by maternal age, parity, ethnicity, and SES. Standardisation is needed

because a region with, e.g., a high number of non-Western women or a high number of

teenage pregnancies will generally have a higher prevalence of adverse perinatal outcomes.

Tables 6.2 and 6.3 show the socio-demographic parameters and the specific criteria for the

PCC and the antenatal risk assessment sub-studies. For each specific indicator we present

Figure 6.2 Absolute prevalence of ‘Big4’ morbidities per 1,000 births.

91

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Ta

ble

6.2

Se

lect

ion

crite

ria*

for t

he p

reco

ncep

tion

care

exp

erim

ent w

ith s

corin

g in

dec

iles;

the

high

er d

ecile

s re

pres

ent a

mor

e lik

ely

qual

ifica

tion

for i

nclu

sion

  

Dem

ogra

phic

s1s

t ant

enat

al b

ooki

ng

≥14W

Cong

enita

l ano

mal

ies

SGA

Rank

#%

PR

EGAG

E <2

0N

W

ETH

NLO

W

SES

ABS

STN

DIN

EQA

BSST

ND

INEQ

ABS

STN

DIN

EQ 

Cit

ies

1A

mst

erda

m10

810

1010

103

32

78

69

962

Rott

erda

m10

1010

1010

103

67

410

96

105

3D

en H

aag

910

1010

1010

29

84

109

810

94

Utr

echt

93

106

1010

410

107

32

791

5Ei

ndho

ven

87

97

99

68

98

99

510

36

Tilb

urg

88

99

54

104

45

1010

389

7A

lmer

e8

710

310

91

77

69

88

938

Gro

ning

en c

ity7

95

92

25

22

45

35

609

Bred

a7

66

53

19

99

36

74

7510

Nijm

egen

75

69

33

94

56

88

679

11En

sche

de6

88

104

42

55

39

76

7712

Ape

ldoo

rn6

53

26

74

11

95

410

6313

Haa

rlem

73

76

87

31

27

44

766

14A

rnhe

m6

99

88

57

66

37

75

8615

Zaan

stad

64

86

77

12

32

54

863

16A

mer

sfoo

rt7

27

49

97

56

63

24

7117

Haa

rlem

mer

mee

r7

14

14

54

11

22

27

4118

's-H

erto

genb

osch

53

34

12

109

94

88

470

19Zo

eter

mee

r5

68

31

16

44

17

610

6220

Zwol

le6

73

42

37

21

72

110

5521

Maa

stric

ht4

94

64

310

1010

1010

101

91

Tabl

e 6.

2 co

ntin

ues o

n ne

xt p

age.

92

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Ta

ble

6.2

Co

ntin

ued

  

Dem

ogra

phic

s1s

t ant

enat

al b

ooki

ng

≥14W

Cong

enita

l ano

mal

ies

SGA

Rank

#%

PR

EGAG

E <2

0N

W

ETH

NLO

W

SES

ABS

STN

DIN

EQA

BSST

ND

INEQ

ABS

STN

DIN

EQ 

22D

ordr

echt

610

97

98

32

13

77

880

23Le

iden

54

76

88

68

710

65

383

24Em

men

46

110

45

102

27

64

970

25Ed

e5

63

55

66

78

21

15

6026

Venl

o3

78

73

28

66

59

101

7527

Wes

tland

41

11

57

610

108

11

1065

28D

even

ter

56

68

78

77

72

77

279

29D

elft

37

99

75

710

1010

55

693

30Si

ttar

d-G

elee

n3

83

71

210

54

49

81

6531

Leeu

war

den

410

49

54

88

82

53

1080

32A

lkm

aar

44

65

66

85

510

22

265

33H

eerle

n2

105

103

48

1010

310

102

8734

Hel

mon

d5

57

66

54

88

59

101

7935

Hilv

ersu

m1

55

39

91

11

83

51

5236

Súdw

est F

rysl

ân3

51

82

29

22

101

13

4937

Am

stel

veen

21

82

88

11

110

21

1055

38H

enge

lo4

64

75

61

43

14

45

5439

Purm

eren

d2

46

49

101

35

14

69

6440

Roos

enda

al2

59

12

18

99

88

101

7341

Oss

22

43

11

75

49

1010

260

42Sc

hied

am1

1010

1010

102

76

410

97

9643

Spijk

enis

se1

97

43

25

33

16

94

57

93

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

44Le

idsc

hend

am-V

oorb

urg

22

73

87

55

49

35

565

45A

lphe

n a/

d Ri

jn1

25

14

49

78

14

46

5646

Alm

elo

38

58

23

11

19

76

155

47Vl

aard

inge

n1

810

57

48

65

98

84

8348

Gou

da3

38

83

19

13

34

33

5249

Mid

delb

urg

19

47

66

48

66

43

367

50Vl

issi

ngen

110

65

86

56

81

89

376

Pro

vin

ces

  

  

 

51G

roni

ngen

87

29

79

53

28

56

778

52Fr

iesl

and

94

18

99

310

108

23

985

53D

rent

he9

31

56

86

44

23

58

6454

Ove

rijss

el9

11

25

72

33

61

29

5155

Gel

derla

nd10

22

21

33

109

92

36

6256

Utr

echt

101

31

23

59

95

11

757

57N

oord

-Hol

land

101

22

78

26

65

11

859

58Zu

id-H

olla

nd10

22

14

54

87

71

29

6259

Zeel

and

83

13

1010

24

51

35

459

60N

oord

-Bra

bant

101

21

11

97

75

67

259

61Li

mbu

rg9

42

21

110

910

67

82

7162

Flev

olan

d8

95

46

66

33

96

67

78

10%

HIG

EST

SCO

RE

10%

-20%

H

IGH

EST

SCO

RE

10%

LO

WES

T SC

ORE

* ‘%

PRE

G’:

% p

regn

ant w

omen

in th

e ge

nera

l pop

ulat

ion

/ ‘AG

E <2

0’: %

teen

age

preg

nanc

ies

/ ‘N

W E

THN

’: %

non

-Wes

tern

pre

gnan

t wom

en /

‘LO

W S

ES’:

% w

omen

in

neig

hbou

rhoo

ds w

ith a

soci

oeco

nom

ic st

atus

scor

e <p

20 /

‘ABS

’: A

bsol

ute

% /

‘STN

D’:

Stan

dard

ised

% /

‘INEQ

’: In

equa

lity

as m

easu

red

by th

e re

lativ

e ris

k of

pre

vale

nces

be

twee

n w

omen

from

nei

ghbo

urho

ods

with

soc

ioec

onom

ic s

tatu

s sc

ore

<p20

com

pare

d to

>p8

0.

94

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Ta

ble

6.3

Se

lect

ion

crite

ria* f

or th

e ris

k se

lect

ion

expe

rimen

t with

scor

ing

in d

ecile

s; th

e hi

gher

dec

iles r

epre

sent

a m

ore

likel

y qu

alifi

catio

n fo

r inc

lusi

on

  

Dem

ogra

phic

sPe

rinat

al m

orta

lity

/ al

l wom

enPe

rinat

al m

orta

lity

/ BI

G4

mor

bidi

ties

Perin

atal

mor

talit

y /

star

t lab

our i

n pr

imar

y ca

re

Rank

#%

PR

EGAG

E <2

0PR

IMI

NW

ET

HN

LOW

SE

SA

BSST

ND

INEQ

ABS

STN

DIN

EQA

BSST

ND

INEQ

 

Cit

ies

1A

mst

erda

m10

810

1010

86

98

77

75

811

32

Rott

erda

m10

107

1010

1010

36

73

109

511

03

Den

Haa

g9

107

1010

98

76

74

108

911

44

Utr

echt

93

910

69

92

910

27

65

965

Eind

hove

n8

79

97

55

42

22

98

683

6Ti

lbur

g8

87

99

88

64

58

99

310

17

Alm

ere

87

410

38

103

58

36

77

898

Gro

ning

en7

910

59

79

18

93

21

787

9Br

eda

76

66

53

47

24

67

83

7410

Nijm

egen

75

86

910

104

1010

26

67

100

11En

sche

de6

85

810

99

48

63

98

396

12A

peld

oorn

65

43

28

88

98

83

410

8613

Haa

rlem

73

97

64

68

56

93

27

8214

Arn

hem

69

109

89

49

96

85

28

102

15Za

anst

ad6

46

86

21

12

14

56

456

16A

mer

sfoo

rt7

26

74

1010

510

107

11

888

17H

aarle

mm

erm

eer

71

54

14

37

76

71

16

6018

's-H

erto

genb

osch

53

103

46

53

44

56

75

7019

Zoet

erm

eer

56

68

31

12

11

16

710

5820

Zwol

le6

76

34

62

58

42

41

1068

21M

aast

richt

49

84

68

78

32

610

96

9022

Dor

drec

ht6

104

97

21

32

15

74

1071

95

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

23Le

iden

54

107

64

29

32

94

53

7324

Emm

en4

64

110

22

13

31

86

1061

25Ed

e5

61

35

74

99

510

13

270

26Ve

nlo

37

58

73

210

31

1010

102

8127

Wes

tland

41

11

11

28

11

88

79

5328

Dev

ente

r5

66

68

99

37

54

910

390

29D

elft

37

89

91

15

11

110

108

7430

Sitt

ard-

Gel

een

38

93

73

17

11

99

91

7131

Leeu

war

den

410

94

95

510

55

105

55

9132

Alk

maa

r4

47

65

22

104

310

34

165

33H

eerle

n2

1010

510

78

61

28

1010

493

34H

elm

ond

55

47

65

48

43

108

82

7935

Hilv

ersu

m1

510

53

75

210

86

33

472

36Sú

dwes

t Fry

slân

35

21

87

710

1010

101

17

8237

Am

stel

veen

21

38

21

110

75

91

110

6138

Hen

gelo

46

34

75

75

66

74

44

7239

Purm

eren

d2

48

64

23

95

49

79

577

40Ro

osen

daal

25

59

12

52

12

59

101

5941

Oss

22

54

33

47

12

78

76

6142

Schi

edam

110

910

1010

109

64

85

14

9743

Spijk

enis

se1

98

74

108

69

84

67

794

44Le

idsc

hend

am-V

oorb

urg

22

77

31

110

43

102

38

6345

Alp

hen

a/d

Rijn

12

85

110

101

1010

54

35

7546

Alm

elo

38

35

81

32

25

16

61

5447

Vlaa

rdin

gen

18

710

57

103

610

310

102

9248

Gou

da3

31

88

63

107

69

22

371

49M

idde

lbur

g1

91

47

13

13

42

22

242

50Vl

issi

ngen

110

46

56

91

47

18

101

73

Tabl

e 6.

3 co

ntin

ues o

n ne

xt p

age.

96

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Ta

ble

6.3

Co

ntin

ued

  

Dem

ogra

phic

sPe

rinat

al m

orta

lity

/ al

l wom

enPe

rinat

al m

orta

lity

/ BI

G4

mor

bidi

ties

Perin

atal

mor

talit

y /

star

t lab

our i

n pr

imar

y ca

re

Rank

#%

PR

EGAG

E <2

0PR

IMI

NW

ET

HN

LOW

SE

SA

BSST

ND

INEQ

ABS

STN

DIN

EQA

BSST

ND

INEQ

 

#P

rov

ince

s 

  

  

  

  

51G

roni

ngen

87

32

99

86

109

65

54

9152

Frie

slan

d9

42

18

109

59

94

46

989

53D

rent

he9

32

15

66

28

82

45

970

54O

verij

ssel

91

11

25

74

89

11

39

6155

Gel

derla

nd10

21

22

56

45

74

34

661

56U

trec

ht10

12

31

45

46

73

34

659

57N

oord

-Hol

land

101

32

24

67

79

61

28

6858

Zuid

-Hol

land

102

22

14

61

58

12

29

5559

Zeel

and

83

21

38

78

1010

52

31

7160

Noo

rd-B

raba

nt10

13

21

33

63

37

78

259

61Li

mbu

rg9

45

22

34

52

36

89

163

62Fl

evol

and

89

15

46

76

79

55

510

87

10%

HIG

EST

SCO

RE

10%

-20%

H

IGH

EST

SCO

RE

10%

LO

WES

T SC

ORE

* ‘%

PRE

G’:

% p

regn

ant w

omen

in th

e ge

nera

l pop

ulat

ion

/ ‘AG

E <2

0’: %

teen

age

preg

nanc

ies

/ ‘PR

IMI’:

% p

rimip

arou

s w

omen

/ ‘N

W E

THN

’: %

non

-Wes

tern

pre

gnan

t w

omen

/ ‘L

OW

SES

’: %

wom

en in

nei

ghbo

urho

ods

with

a s

ocio

econ

omic

sta

tus

scor

e <p

20 /

‘ABS

’: A

bsol

ute

% /

‘STN

D’:

Stan

dard

ised

% /

‘INEQ

’: In

equa

lity

as m

easu

red

by th

e re

lativ

e ris

k of

pre

vale

nces

bet

wee

n w

omen

from

nei

ghbo

urho

ods

with

soc

ioec

onom

ic s

tatu

s sc

ore

<p20

com

pare

d to

>p8

0.

97

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

the absolute rate (ABS), the standardised rate (STND) and the inequality-rate (INEQ), the

latter being expressed as the relative risk of the outcome for low SES pregnant women

compared to high SES pregnant women, after direct standardisation for maternal age,

parity and ethnicity.17 Next, to facilitate comparisons, we assigned decile scores to regions,

varying from one (the region is one of the 10% areas with best outcomes) to 10 (the region

belongs to the 10% worst outcomes). The sum of the decile scores for the various indicators

by region is shown in the last column (‘RANK’); higher scores imply unfavourable ranking.

For clarity, colors are used to differentiate between favourable and unfavourable outcomes:

green represents the first decile (with the best outcome), pink the 10th decile (10% with

the most adverse outcomes), amber the 10th-20th decile. Based on the sum of the decile

scores for the PCC sub-study (table 6.2), the pink and amber municipalities/regions have

the most adverse outcomes, i.e., 1. The Hague; 2. Rotterdam; 3. Eindhoven; 4. Amsterdam; 5.

Schiedam; 6. Almere ; 7. Delft; 8. Utrecht; 9. Maastricht; 10. Tilburg; 11. Heerlen; 12. Arnhem;

13. Friesland. According to the sum of the decile score for the risk assessment sub-study

(table 6.3) the following municipalities/regions show the most adverse outcomes: 1. The

Hague; 2. Amsterdam; 3. Rotterdam; 4. Arnhem; 5. Tilburg; 6. Nijmegen; 7. Schiedam; 8.

Utrecht; 9. Enschede; 10. Spijkenisse; 11. Heerlen; 12. Vlaardingen; 13. Groningen; 14.

Leeuwarden.

Additional to the identified municipalities, the province of Friesland best qualified for the

PCC sub-study and the province of Groningen for the risk assessment sub-study.

Final selection municipalities/regions

After the theoretical selection of the candidate municipalities/regions in the previous step,

the next step elaborated on the practical aspects of sub-study implementation in candidate

municipalities/regions, e.g., willingness to participate of local authorities and caregivers.

The candidate municipalities/regions were first presented to the Ministry of Health, after

which the Aldermen of these candidate municipalities/regions were consulted, which

resulted in a final selection. The participating municipalities/regions are (figure 6.3): in the

province of Groningen Appingedam/Delfzijl/Menterwolde/Pekela and Groningen city, the

municipalities of Enschede, Nijmegen, Heerlen, Tilburg, Schiedam, Utrecht, The Hague,

Amsterdam, and Almere.

All municipalities decided to participate in both sub-studies either as study or control. As a

separate municipal program on reducing perinatal mortality was already being carried out

in Rotterdam5, this city was not selected for participation in the HP4All study.

98

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

DESIGN OF THE PRECONCEPTION CARE SUBSTUDYIn this population-based prospective cohort study with a pre-post design for the

measurement of effectiveness of PCC, we approach all women in the fertile age (18-42

years) and invite them to visit Preconception Care Consultations if they contemplate

pregnancy. This sub-study consists of two consultations provided by midwives or general

practitioners (GP), the second consultation takes place three months after the first. During

Figure 6.3 Participating municipalities in the ‘Healthy Pregnancy 4 All’ project.

AppingedamDelfzijl

GroningenMenterwolde

Pekela

Enschede

Heerlen

Nijmegen

Tilburg

Schiedam

Utrecht

AlmereAmsterdam

The Hague

99

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

the first consultation, PCC risks are assessed and a tailored management plan is composed.

At the second consultation, reduction of the PCC risk load is measured, and if required, the

management plan is adapted.

Sample size calculation

The statistical design is based on a pre-post comparison with paired data. The primary

outcomes used for the sample size calculation are overall preconceptional folic acid

supplementation use and smoking cessation amongst smokers, as changes in these

behavioral risk factors have a major impact on perinatal health and can be measured by

biomarkers.22-24

The sample size was based on the following criteria:

1. Self-reported folic acid supplementation use. To reject the null hypothesis (H0: Δ ≤ 20%,

defined as the PCC program leading to a ≤ 20% increase of folic acid users in women

that were not already using folic acid supplements at baseline) a total sample size of

n= 839 is needed. Assumptions for the power calculation were (1) the smallest clinically

relevant difference (‘Δ’) is a 20% increase of folic acid use in non-users at baseline, (2) the

proportion of women using folic acid at baseline is 30% (π0=30%), (3) a-select drop-out

rate of 10%, (4) results are pairwise analysed, (5) a statistical significance level α <0.025

(1-sided, correction for multiple testing due to two primary outcome measures), and

(6) a power (1-β) of 0.80.

2. Smoking cessation. (1) To reject the null hypothesis (H0: Δ ≤ 5%, defined as smoking

cessation occurring in ≤ 5% of women that smoked at baseline) a total sample size of

n=687 is needed. Assumptions for the power calculation were (2) the smallest clinically

relevant difference (‘Δ’) is a 5% decrease of smoking compared to baseline, (3) the

proportion of smoking women at baseline is 30% (π0=30%), (4) the a-select drop-out

rate is 10%, (5) the results are pairwise analysed, (5) the significance level is α <0.025

(1-sided, correction for multiple testing due to two primary outcome measures), and

(6) the power (1-β) is 0.80.

Enrollment

Women are actively approached by: (1) an invitational letter from the municipal public health

service or municipality, and/or (2) an invitational letter from their GP, and/or (3) referral by

a youth healthcare physician or nurse, and/or (4) referral by a peer educator in perinatal

100

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

health. The precise mode of approach depends on specific local collaborative agreements

with the municipal authorities, the local public health authorities and the local caregivers.

In addition, participating GPs and midwives can recruit women regularly attending their

care. Furthermore, women within the community are informed about the PCC consultations

by flyers and posters. Each woman, including the method that they were recruited by, is

registered when a PCC visit at participating GP or midwife practices is scheduled. Women

are sent participant information leaflets with an informed consent form before they are

approached by telephone for inclusion. If women agree to participate they hand in the

informed consent form before their consultation.

Logistics and data collection

The logistics of the PCC sub-study are carried out in close collaboration with local project

coordinators and certified clinical laboratories in the 14 participating municipalities. The

PCC risk assessment is performed by the client prior to each consultation individually at a

convenient moment using the web-based validated ‘ZwangerWijzer’ (translated ‘Preparing

for pregancy’) internet questionnaire (www.zwangerwijzer.nl).25 GPs and midwives during

the consultations are supported professionally by the ‘Preconceptiewijzer’ (translated

‘Preparing for preconception care’) tool (www.preconceptiewijzer.nl)., ‘Preconceptiewijzer’

presents identified risk factors from ‘ZwangerWijzer’.This tool further provides GPs and

midwives both with protocols on management of PCC risk factors and with information

leaflets to hand out to the client.

Before the first PCC visit takes place, additional information will be collected from partici-

pating women: data on basic characteristics (e.g., educational level), general health (e.g.,

medical history), details on risk factors and health behaviours (primarily: folic acid use,

alcohol consumption, smoking, drug use, diet, preventive behaviours regarding Listeria/

Toxoplasmosis prevention, weight) , and attitudes towards PCC are collected by a digital

or on request by a paper - questionnaire (table 6.4). This questionnaire also includes

a cardiovascular risk score (My Life Check, American Heart association, online: http://

mylifecheck.heart.org/) and a risk score for diabetes (FINDRISK).26 To measure alterations

after the first PC consultation, clients are sent a digital or paper questionnaire regarding

their health behaviours (risk reduction or elimination) 3 months after the first consultation.

The same health behaviours are addressed as in the first questionnaire.

Additionally, protocol-based anthropometric measurements are performed (BMI, blood

pressure, waist- and hip circumference), and biomarkers are collected at both consultations.

101

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Ta

ble

6.4

Pl

anne

d as

sess

men

ts in

the

prec

once

ptio

n ca

re e

xper

imen

t: pa

ram

eter

s, bi

omar

kers

, que

stio

nnai

res

and

outc

ome

Para

met

er

Bio

mar

ker

Que

stio

nnai

re O

utco

me

Folic

aci

d su

pple

tion

Eryt

rocy

te fo

late

, aft

er b

oth

cons

ulta

tions

Self-

repo

rted

folic

aci

d us

e•

Prev

alen

ce: f

olic

aci

d us

e m

easu

red

at b

asel

ine

with

ery

troc

yte

fola

te a

nd s

elf-r

epor

ted

use.

• Ef

fect

iven

ess

of p

reco

ncep

tion

care

: inc

reas

e in

num

ber o

f wom

en

usin

g fo

lic a

cid

(sel

f-rep

orte

d an

d in

crea

se o

f ery

troc

yte

fola

te

com

pare

d to

bas

elin

e).

Smok

ing

cess

atio

nSe

rum

cot

inin

e le

vel,

afte

r bo

th c

onsu

ltatio

nsSe

lf-re

port

ed s

mok

ing,

sm

okin

g ce

ssat

ion

and

smok

ing

redu

ctio

n.

• Pr

eval

ence

: num

ber o

f sm

oker

s at

bas

elin

e ba

sed

on s

elf-r

epor

t an

d se

rum

cot

inin

e.

• Ef

fect

iven

ess

prec

once

ptio

n ca

re: n

umbe

r of w

omen

repo

rtin

g to

ha

ve s

topp

ed s

mok

ing

or h

ave

redu

ced

smok

ing.

Alc

ohol

inta

ke c

essa

tion

% C

DTa a

nd G

amm

a G

T,

afte

r bot

h co

nsul

tatio

nsSe

lf-re

port

ed a

lcoh

ol

cons

umpt

ion,

bef

ore

each

co

nsul

tatio

n.

• Pr

eval

ence

: num

ber o

f wom

en u

sing

alc

ohol

and

hav

e be

en b

inge

dr

inki

ng, o

r cla

ssifi

ed a

s he

avy

drin

kers

(sel

f-rep

orte

d us

e),%

CD

T an

d G

amm

a G

T at

bas

elin

e.

• Ef

fect

iven

ess

of p

reco

ncep

tion

care

: cha

nge

in n

umbe

r of w

omen

us

ing

alco

hol p

reco

ncep

tiona

lly (s

elf-r

epor

t). D

ecre

ase

in %

CD

T an

d G

amm

a G

T, n

orm

aliz

ing

for h

eavy

drin

kers

aft

er a

bstin

ence

/de

crea

sed

use.

Dru

gsTo

x sc

reen

ing

urin

eb , aft

er

both

con

sulta

tions

Que

stio

nnai

re a

nd m

easu

re

befo

re e

ach

cons

ulta

tion

• Pr

eval

ence

: num

ber o

f wom

en u

sing

har

d- a

nd s

oftd

rugs

at

base

line

base

d on

sel

f-rep

ort a

nd u

rine

scre

enin

g.•

Effe

ctiv

enes

s of

pre

conc

eptio

n ca

re: n

umbe

r of w

omen

who

st

oppe

d dr

ug u

se b

ased

on

self-

repo

rt a

nd u

rine

scre

enin

g

Tabl

e 6.

4 co

ntin

ues o

n ne

xt p

age.

102

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Ta

ble

6.4

Co

ntin

ued

Para

met

er

Bio

mar

ker

Que

stio

nnai

re O

utco

me

Vita

min

e D

25-O

H C

alci

diol

, aft

er th

e fir

st c

onsu

ltatio

n –

•

Prev

alen

ce o

f vita

min

D d

efic

ienc

y ba

sed

on 2

5-O

H C

alci

diol

leve

l m

easu

red

at th

e st

art o

f the

pre

conc

eptio

n ca

re c

onsu

ltatio

n.•

Que

stio

ns re

gard

ing

prog

nost

ic fa

ctor

s

Nut

ritio

nN

ot a

pplic

able

Self

repo

rted

die

t, be

fore

eac

h co

nsul

tatio

n•

Avoi

danc

e of

vita

min

A, n

o ra

w p

rodu

cts

and

prop

erly

was

hed

vege

tabl

es: s

elf-r

epor

ted

at b

asel

ine.

• In

crea

se in

num

ber o

f wom

en w

ith a

hea

lthy

diet

, avo

idin

g vi

tam

in A

and

raw

pro

duct

s, ve

geta

bles

pro

perly

was

hed:

sel

f-re

port

ed d

urin

g se

cond

con

sulta

tion

com

pare

d to

bas

elin

e.

Adju

stm

ent m

edic

atio

n re

gim

eN

ot a

pplic

able

Self

repo

rted

med

icat

ion

use,

be

fore

eac

h co

nsul

tatio

n.•

Prev

alen

ce: p

reco

ncep

tiona

l med

icat

ion

use

and

type

s of

m

edic

atio

n.•

Effe

ctiv

enes

s: s

afe

med

icat

ion

polic

y ev

alua

tion

by th

e he

alth

car

e pr

ovid

er ([

1] re

ferr

al fr

om m

idw

ife to

GP

for m

edic

atio

n po

licy,

[2]

num

ber o

f cas

es in

whi

ch c

essa

tion/

repl

acem

ent/

no c

essa

tion/

cont

inua

tion

of m

edic

atio

n is

cho

sen,

bas

ed o

n 'T

erat

olog

y In

form

atio

n Se

rvic

e' a

dvic

e, [3

] pat

ient

adh

eren

ce to

med

icat

ion

polic

y

Purs

uing

a h

ealth

y bo

dyw

eigh

t–

Ant

hrop

omet

rics

in b

oth

cons

ulta

tions

•

Prev

alen

ce o

f ove

rwei

ght (

BMI 2

5-30

), ob

esity

(BM

I >30

), an

d un

derw

eigh

t (BM

I <18

.5)

• Ef

fect

iven

ess

of p

reco

ncep

tion

care

: num

ber o

f wom

en w

ith

over

wei

ght B

MI >

25 lo

sing

wei

ght,

and

unde

rwei

ght B

MI <

18.5

%

gain

ing

wei

ght.

103

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Adhe

renc

e to

pe

rson

aliz

ed a

dvic

e–

• Ad

here

nce

to a

dvic

e re

gard

ing

spec

ific

dete

cted

risk

fact

ors

acco

rdin

g to

the

care

give

r.

Preg

nanc

y ou

tcom

e–

Retr

ospe

ctiv

e fr

om n

atio

nal

data

base

(The

Net

herla

nds

Perin

atal

Reg

istr

y)

• N

umbe

r of b

irths

adv

erse

out

com

es d

efin

ed a

s (1

) per

inat

al

mor

talit

y, (2

) con

geni

tal a

nom

alie

s, (3

) pre

term

birt

h (<

37th

wee

k of

pre

gnan

cy),

(4) s

mal

l for

ges

tatio

nal a

ge (b

irthw

eigh

t <10

th

perc

entil

e), (

5) lo

w A

pgar

sco

re (<

5, 5

min

utes

aft

er b

irth)

.

Book

ing

visi

t <12

th w

eek

of g

esta

tion

– Re

tros

pect

ive

from

The

N

ethe

rland

s Pe

rinat

al

Regi

stry

, sel

f-rep

orte

d at

firs

t co

nsul

tatio

n

• G

esta

tiona

l age

at f

irst a

nten

atal

boo

king

vis

it

Card

iova

scul

ar ri

sk s

core

ac

cord

ing

to 'M

y Li

fe

Chec

k'c

Chol

este

rol,

Glu

cose

Mea

sure

d bl

ood

pres

sure

, BM

I, qu

estio

nnai

re it

ems

rega

rdin

g di

etar

y ha

bits

and

phy

sica

l ac

tivity

• Pr

ogno

stic

val

ue o

f car

diov

ascu

lar r

isk

scor

e an

d pr

eval

ence

of:

hype

rten

sive

dis

ease

in p

regn

ancy

, con

geni

tal a

nom

alie

s, pr

eter

m

birt

h, s

mal

l for

ges

tatio

nal a

ge, l

ow A

pgar

sco

re.

FIN

DRI

SK d

iabe

tes

scor

eCh

oles

tero

l, G

luco

seBl

ood

pres

sure

, BM

I, w

aist

an

d hi

p ci

rcum

fere

nce,

qu

estio

nnai

re it

ems

rega

rdin

g fa

mily

his

tory

, abn

orm

al

gluc

ose

mea

sure

men

t

• Pr

ogno

stic

val

ue d

iabe

tes

risk

scor

e an

d pr

eval

ence

of:

gest

atio

nal

diab

etes

, hyp

erte

nsiv

e di

seas

e in

pre

gnan

cy, p

erin

atal

mor

talit

y,

mac

roso

mia

, con

geni

tal a

nom

alie

s, pr

eter

m b

irth,

sm

all f

or

gest

atio

nal a

ge, l

ow A

pgar

sco

re.

a CD

T: C

arbo

hydr

ate-

defic

ient

tran

sfer

rin, l

abor

ator

y te

st to

det

ect (

heav

y) a

lcho

hol c

onsu

mpt

ion.

b scr

eens

for a

mfe

tam

ines

, bar

bitu

rate

s, be

nzod

iaze

pine

s, c

ocai

ne, e

csta

sy,

met

amfe

tam

ine,

met

hado

n, o

piat

es, P

CP, c

anna

bis.

c htt

p://

myl

ifech

eck.

hear

t.org

.

104

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Blood samples are collected either by the caregiver or by an outpatient-testing facility of

the local laboratory. Initial analysis of the samples (hemoglobin, mean corpuscular volume,

fasting glucose at baseline) and refrigeration (between -20 and -80 degrees Celsius) will

occur within 24 hours after sample collection at the local laboratory. Only deviant vitamin

D, glucose, cholesterol and hemoglobin results are communicated to the participant’s GP if

she has requested so in the informed consent form. Regarding primary outcomes: erytrocyte

folate – as biomarker for folic acid use), serum cotinin – as biomarker for cigarette smoking,

carbohydrate dehydrogenase (CDT) – as biomarker for outcome for heavy drinking and

urinary drugs tests – as biomarker for soft and hard-drug use; are measured.

Study preparation

Before study onset, the participating midwives and GPs receive training in PCC, including

implementation of the ‘Preconceptiewijzer’ tool. Additionally, the participating caregivers

and their staff are given a preparatory tutorial at baseline on relevant procedures within the

HP4All study. These procedures include anthropometric and blood pressure measurements

and paper and web-based data registration.

Recruitment for the PCC sub-study was set up to the extent that was agreed upon with

the local project coordinators. Firstly, the municipality or Municipal Health Service (which

ever was chosen locally) selected women by the municipal population registry to mail

the invitational letter. An invitational letter was provided by the project, and letters were

adapted locally to fit the local situation. Secondly, local youth service was provided with

flyers and posters- with local practices offering consultations - to disseminate amongst their

youth healthcare physicians and nurses. Furthermore, GPs were supported logistically in

selecting women registered in their practice. Women with evident contraindications for

such invitation, both medically (such as terminal illness, with a hysterectomy or sterilisation

in the past) or socially (widow, in detention) were excluded. They sent out a locally adapted

basic letter provided by the project. Perinatal Health Educators were trained and provide

perinatal health education sessions to self-recruited women contemplating pregnancy.

DESIGN OF THE ANTENATAL RISK ASSESSMENT SUBSTUDYIn this cluster randomised trial midwifery practices in participating municipalities (‘clusters’)

were randomly assigned to either the use of a score-card (‘R4U’) based antenatal risk

105

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

assessment, care pathways and multidisciplinary consultation (intervention group) or

conventional risk assessment (control group). Score-card based systematic risk assessment

will be performed with the ‘R4U’ score-card at the first antenatal booking visit followed

by, if necessary, a specific referral to, e.g., higher level obstetric care (gynaecologist),

psychosocial care in case of medical or non-medical high risk using risk-specific care

pathways. Additionally, these women at increased risk will be reviewed in a multidisciplinary

team of caregivers concerning tailored antenatal care. Exclusion criteria include a medical

emergency situation during the booking visit (e.g., ectopic pregnancy), or women being

in labour.

The 70-item ‘R4U’ score-card consists of six risk domains (social status, ethnicity, care, lifestyle,

medical history and obstetric history). Corresponding care pathways to both medical and

non-medical services will support health care professionals to encounter complex (non-)

medical risk factors. A predefined weighted sum risk threshold, based on weighted single

risk factors, is derived from the ‘R4U’ score-card. If a pregnant woman’s individual sum risk

score exceeds the threshold, her case will be assessed in a multidisciplinary setting with

community midwives, obstetricians, and other care providers.

Pregnant women’s risk status in the control group is assessed conventionally, i.e., according

to the elaborate so-called ‘List of Obstetric Indications’ (in Dutch: Verloskundige Indicatie

Lijst)12 which lists all conventional (>140) high risk indications (for referral or consultation). In

each control region care ‘as usual’ will be provided until 700 participants have been included

or until 2/3 of the study period (2 years) has passed. After that moment, the implementation

of the risk assessment intervention will start.

Primary outcomes are the prevalence of preterm birth and SGA, and the efficacy of ‘R4U’

implementation (measured by the number of ‘R4U’ score-cards completed by the health

care professional against the number of booking visits, the development and use of care

pathways following ‘R4U’ scores, actually performed multidisciplinary consultations,

and patient and healthcare professional satisfaction). Secondary outcomes are perinatal

mortality (from 22 weeks’ gestation until 7 days postpartum), undetected SGA and

unexpected preterm birth, measured as onset of childbirth in primary care, prevalence and

accumulation of medical and non-medical risk factors, and the number of women referred

to secondary care through either R4U score-card based or conventional risk assessment.

106

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Randomisation procedure and sample size calculation

In January 2011 randomisation took place by an independent statistician. In order to study the

need for matching, we stratified the adverse perinatal outcomes (‘Big4’ and perinatal mortality)

in each cluster by SES and ethnicity. Since clusters were not considerably different in terms of

these characteristics, matching of clusters for SES and ethnicity was considered unnecessary.

Sample size in this study is based on the effectiveness of the ‘R4U’ risk assessment and

resulting care pathways intervention in early pregnancy on the likelihood of SGA and

preterm birth. In this cluster-randomised trial, sample size depends on: (1) the average

risk of SGA and preterm birth without the intervention (π0); (2) the expected effect of the

intervention (π1); (3) the inflation factor reflecting the partial similarity/dependency of

women’s outcomes or responses within the same cluster27); and (4) the α and (5) power (1-β)

of the test. Using data from the 2000-2008 Netherlands Perinatal Registry for the selected

postal code regions20, we estimated π0 at 16.7% (summation of preterm birth and SGA),

the expected effect of the intervention (π1) at 13% and the inflation factor at 2.06, based

on formulas on Donner et al.27 With 10 clusters, five municipalities in the control and five

in the intervention group, and an alpha of 0.05 (two-sided), 7,000 participants (3,500 per

arm) should provide power in excess of 80%. This means that 700 women per cluster should

be included in the analysis.

Logistics and data collection

As with the PCC sub-study, the logistics of this sub-study are carried out in close collaboration

with participating local project coordinators, midwives, obstetricians, and, if available, research

midwives in the 14 participating municipalities. Midwives and obstetricians inform pregnant

women about the HP4All study at the booking visit and hand out an information package.

All pregnant women at their first antenatal booking visit with a midwife or obstetrician

located in the areas are eligible if she meets the inclusion criteria. In intervention

municipalities, the ‘R4U’ score-card will be filled out during the first antenatal visit. For the

‘R4U’ score-card a web-based registration form is used in which women are registered, coded

with a study identification number. Depending on the detected risk factors, care pathways

can be used to modify these risk factors. If a pregnant woman’s weighted sum risk score

exceeds the predefined threshold, multidisciplinary consultation will be advised. In control

municipalities, pregnant women in the control group will receive regular antenatal health

care. Table 6.5 provides an overview of the planned assessments within the risk assessment

sub-study including variables, methods and outcomes.

107

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Table 6.5 Planned assessments in the risk selection experiment: variables, methods and outcomes

Variables Methods Outcomes

Patients

1. INTERVENTION GROUP

Non-medical risk factors: 39 items from the risk score card, categorized into the domains social, ethnicity, care, and lifestyle.

Medical risk factors: 30 items from the risk score card, categorized into the domains general history and obstetric history

Baseline characteristics: Age, zip-code, ethnicity, onset of care, household composition, family income, employment, education level, smoking, alcohol, drugs, folic acid use, medication use, pre-existing chronic diseases, sexually transmitted diseases.

R4U score card + registration form ‘Obstetric history’

Questionnaire ‘Baseline characteristics’

Case Record Form ‘pregnancy and delivery data’

Primary outcomes: - Preterm birth - SGA

Secondary outcomes: - Undetected SGA and unexpected

preterm births (babies born in primary care)

- Prevalence of risk factors- Risk accumulation - Involved healthcare professionals

during pregnancy- Detection and prevention of impaired

growth and preterm birth during pregnancy

- Perinatal mortality - Congenital anomalies- Mode of delivery - Place of delivery - Asphyxia- Neonatal admission - Maternal morbidity (e.g., pre-

existing chronic disease, pregnancy complications), and maternal mortality.

2. CONTROL GROUP

Baseline characteristics: Age, zip-code, ethnicity, onset of care, household composition, family income, employment, education level, smoking, alcohol, drugs, folic acid use, medication use, pre-existing chronic diseases, sexually transmitted diseases.

Registration form ‘Obstetric history’

Questionnaire ‘Baseline characteristics’

Case Record Form ‘pregnancy and delivery data’

Patient satisfaction in both groups Questionnaire ‘Patient experiences during the first antenatal visit’

- Which topics were discussed (10 examples)?

- What was your experience? - Do you think this was important to ask?

Care providers

General characteristics participating midwifery practices and hospitals in both groups

Interview-based questionnaire

- Current number of patients and employees

- Use of risk selection instruments- Collaboration with hospitals and

(other) midwifery practices- Work processes (e.g., counselling for

prenatal screening, or ultrasound facilities).

Care provider satisfaction in both groups

Questionnaire - Feasibility - Efficacy of implementation- Collaboration - Continuation of intervention

108

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

Study preparation

In municipalities allocated to the intervention group, midwives and obstetricians will use

the ’R4U’ score-card during the first antenatal visit of all women (provided that informed

consent is given). Participating midwives and obstetricians receive personal instructions in

planned sessions by the project team for the practical use of the web-based ’R4U’ score-card.

Besides, an e-learning program is available for all caregivers. The project team has developed

28 templates of care pathways for all risk factors in the ‘R4U’ score-card. Together with local

healthcare professionals in perinatal care, municipal services, community health services,

and other services, these templates will be adapted in organised meetings to local setting,

taking the availability of local facilities, agreements, and guidelines into consideration.

We aim to assess 20% of all pregnant women in this multidisciplinary setting in order to

determine a cut-off score per municipality.

INFORMED CONSENT, DATA QUALITY, CONTROL, MANAGEMENT AND TIME SCHEDULE FOR BOTH SUBSTUDIES

Informed consent

The HP4All study has been approved by the Medical Ethical Committee of the Erasmus

Medical Centre Rotterdam (Preconception Care sub-study: MEC 2012-425; Antenatal risk

assessment trial: MEC 2012-322), and by the management of all participating care providers.

Pregnant and non-pregnant women, depending on the sub-study, will receive written and

oral information about the study.

Participation in either sub-study is voluntary and informed consent must be obtained. PCC

consultations are currently not covered by health care insurances; PCC providers will receive

reimbursement from the HP4All project. The participating practices in the antenatal risk

sub-study receive no financial compensation.

Data quality, control and management

For logistic reason, data are collected and coded with a study identification number.

Otherwise we will not be able to link data from the initial visit to birth records. The research

team and participating caregivers have access to the key that links study identification

109

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

number to record data. Data analysis and storage will take place on an anonymised

dataset. Access to this anonymised data is only available to the research team. All data and

biomaterials are stored 15 years after inclusion.

Time schedule

The HP4All study was initiated in April 2011. Full participant recruitment and complete

data collection started in December 2012. A cohort of 839 women for the PCC sub-study

is expected to be completed by the beginning of 2014. For the risk assessment sub-study,

7,000 women are also expected to be included by the beginning of 2014.

REFERENCES1. Buitendijk SE, Nijhuis JG. High perinatal mortality

in the Netherlands compared to the rest of Europe. Ned Tijdschr Geneeskd. 2004;148:1855-1860.

2. Mohangoo AD, Buitendijk SE, Hukkelhoven CW, et al. Higher perinatal mortality in The Netherlands than in other European countries: the Peristat-II study. Ned Tijdschr Geneeskd. 2008;152:2718-2727.

3. The EURO-PERISTAT project in collaboration with SCPE EE. European perinatal health report, 2008. (online: www.europeristat.com).

4. de Graaf JP, Ravelli AC, de Haan MA, Steegers EA, Bonsel GJ. Living in deprived urban districts increases perinatal health inequalities. J Matern Fetal Neonatal Med. 2013;26:473-481.

5. Denktas S, Bonsel GJ, Van der Weg EJ, et al. An urban perinatal health programme of strategies to improve perinatal health. Matern Child Health J. 2012;16:1553-1558.

6. Saravelos SH, Regan L. The importance of pre-conception counseling and early pregnancy monitoring. Seminars in reproductive medicine. 2011;29:557-568.

7. Barker DJ, Bagby SP, Hanson MA. Mechanisms of disease: in utero programming in the patho-genesis of hypertension. Nat Clin Pract Nephrol. 2006;2:700-707.

8. Mook-Kanamori DO, Steegers EA, Eilers PH, Raat H, Hofman A, Jaddoe VW. Risk factors and outcomes associated with first-trimester fetal growth restriction. JAMA. 2010;303:527-534.

9. Korenbrot CC, Steinberg A, Bender C, Newberry S. Preconception care: a systematic review. Matern Child Health J. 2002;6:75-88.

10. van der Zee B, de Beaufort I, Temel S, de Wert G, Denktas S, Steegers E. Preconception care: an essential preventive strategy to improve children’s and women’s health. J Public Health Policy. 2011;32:367-379.

11. Health Council of the Netherlands. Preconception care: a good beginning. The Hague: Health Council of the Netherlands, 2007; publication no. 2007/19E. ISBN 978-90-5549-678-5 (online: http://www.gezondheidsraad.nl/en/publications/prevention/preconception-care-good-beginning).

12. Amelink-Verburg MP, Buitendijk SE. Pregnancy and labour in the Dutch maternity care system: what is normal? The role division between midwives and obstetricians. J Midwifery Womens Health. 2010;55:216-225.

13. de Graaf JP, Ravelli AC, Wildschut HI, et al. Perinatal outcomes in the four largest cities and in deprived neighbourhoods in The Netherlands. Ned Tijdschr Geneeskd. 2008;152:2734-2740.

110

THE ‘H

EALTH

Y PREGN

AN

CY 4 ALL’ STU

DY

6

14. Poeran J, Maas AFG, Birnie E, Denktas S, Steegers EA, Bonsel GJ. Social deprivation and adverse perinatal outcomes among Western and non-Western pregnant women in a Dutch urban population. Soc Sci Med. 2013 (in press).

15. Jaddoe VW, Mackenbach JP, Moll HA, et al. The Generation R Study: Design and cohort profile. Eur J Epidemiol. 2006;21:475-484.

16. van Eijsden M, Vrijkotte TG, Gemke RJ, van der Wal MF. Cohort profile: the Amsterdam Born Children and their Development (ABCD) study. Int J Epidemiol. 2010;40:1176-1186.

17. de Graaf JP, Steegers EA, Bonsel GJ. Inequalities in perinatal and maternal health. Curr Opin Obstet Gynecol. 2013;25:98-108.

18. Bonsel GJ, Birnie E, Denktas S, Poeran J, Steegers EAP. Dutch report: Lijnen in de Perinatale Sterfte, Signalementstudie Zwangerschap en Geboorte 2010. Rotterdam: Erasmus MC; 2010.

19. van der Kooy J, Poeran J, de Graaf JP, et al. Planned home compared with planned hospital births in the Netherlands: intrapartum and early neonatal death in low-risk pregnancies. Obstet Gynecol. 2011;118:1037-1046.

20. Perinatal care in The Netherlands 2008 (in Dutch: Perinatale zorg in Nederland 2008). Utrecht: The Netherlands Perinatal Registry; 2011.

21. Roalfe AK, Holder RL, Wilson S. Standardisation of rates using logistic regression: a comparison with the direct method. BMC Health Serv Res. 2008;8:275.

22. Andres RL, Day MC. Perinatal complications associated with maternal tobacco use. Semin Neonatol. Aug 2000;5:231-241.

23. Lumley J, Watson L, Watson M, Bower C. Pericon-ceptional supplementation with folate and/or multivitamins for preventing neural tube defects. Cochrane Database Syst Rev. 2001:CD001056.

24. Salmasi G, Grady R, Jones J, McDonald SD. Environmental tobacco smoke exposure and perinatal outcomes: a systematic review and meta-analyses. Acta Obstet Gynecol Scand. 2010;89:423-441.

25. Landkroon AP, de Weerd S, van Vliet-Lachotzki E, Steegers EA. Validation of an internet question-naire for risk assessment in preconception care. Public Health Genomics. 2009;13:89-94.

26. Lindstrom J, Tuomilehto J. The diabetes risk score: a practical tool to predict type 2 diabetes risk. Diabetes care. 2003;26:725-731.

27. Donner A, Klar N. Design and analysis of cluster randomization trials in health research. 1st ed. London; 2000.

7

Manuscript submitted for publication.

The effectiveness of risk selection in the Dutch obstetric care system

J. Poeran, J. van der Kooy, E. Birnie, S. Denktaş, E.A.P. Steegers, G.J. Bonsel

112

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

ABSTRACTObjective To quantify if risk selection in Dutch obstetric care (by midwives) results in a

true low risk population in primary care at the end of pregnancy. This is an essential quality

of care indicator as a distinction is made between primary care for low risk pregnancies by

independently practicing community midwives, and secondary/tertiary care for high risk

pregnancies by obstetricians.

Methods All singleton pregnancies (≥22 weeks’ gestation, 2000-2007, n=1,407,387) from

The Netherlands Perinatal Registry were selected. We defined high risk pregnancy as the

presence of ≥1 ‘Big4’ morbidities, the main precursors of perinatal mortality: congenital

anomalies, preterm birth, small for gestational age (SGA), or low Apgar score. Referral

patterns of high risk pregnancies were studied during pregnancy and parturition; adequate

risk selection implies no high risk pregnancies in primary care. Additionally, we applied a

diagnostic test framework to study effectiveness of SGA selection (and referral) by defining

true positives (referral of SGA), false positives (referral of non-SGA), false negatives (non-

referral of SGA), and true negatives (non-referral of non-SGA). Sensitivity, specificity, negative

predictive value (NPV), negative likelihood ratio (LR-), and false negative rate (FN) were

determined for eight patient subgroups.

Results 59% of ‘Big4’ were referred during pregnancy, 19% during parturition; 22%

remained in primary care. SGA ‘test’ characteristics differed considerably for subgroups

(sensitivity 15%-59%, specificity 54%-87%, NPV 89%-97%, LR- 0.69-1.05, FN 3%-11%).

Conclusion Risk selection in Dutch obstetric care does not realise its aim of a true low risk

group in primary care at the end of pregnancy. Methods for improvement are warranted.

113

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

INTRODUCTIONIn The Netherlands perinatal mortality exceeds the European average.1 The unique Dutch

system of obstetric care has been regarded as a potential contributing factor.2-4 This system

is characterised by three risk-based levels of care. Primary care for low risk pregnancies is

provided by independently practicing community midwives and a small percentage of

general practitioners (GPs). Assumed low risk pregnant women can either opt for a home

birth or a short-stay hospital birth under supervision of a community midwife. Secondary/

tertiary care for assumed high risk pregnancies is provided by obstetricians in hospitals.

Currently, approximately 80% of pregnant women start antenatal care in primary care.5

Whenever risk factors (for adverse perinatal or maternal outcome) are present before

pregnancy or arise during pregnancy or parturition, women shift from low risk to high risk

and are referred to secondary care or from secondary to tertiary care, also during parturition.

This ongoing risk assessment during pregnancy and during parturition is called ‘risk selection’.

In formal terms, the aim of risk selection is to identify and refer high risk pregnancies in order

to obtain a true low risk group of pregnant women (expressed as high negative predictive

value of risk selection) in the primary care setting.5,6 Thus, risk selection adequacy is an

essential quality of care indicator of the Dutch obstetric care system.

Although the effectiveness of risk selection in Dutch primary obstetric care has been studied,

a nationwide systematic evaluation on the performance of the risk selection process is still

absent.3,4,6-11 The present nationwide retrospective study quantifies the performance of risk

selection (during pregnancy and during parturition) by community midwives in terms of its

ability to achieve a true low risk population at the end of pregnancy.

METHODS

Netherlands Perinatal Registry

We selected data from all singleton pregnancies for the period 2000-2007 as registered in

The Netherlands Perinatal Registry, which is subject to Dutch law regulations regarding

confidentiality. In agreement with the World Health Organization (WHO) reporting

guidelines, only pregnancies with a gestational age of ≥22 weeks were included.12

The registry contains population-based information of >97% of all pregnancies in The

Netherlands.13 Source data are collected by 94% of midwives, 99% of gynaecologists and

68% of paediatricians (including 100% of Neonatal Intensive Care Unit paediatricians) as part

114

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

of their routine medical dossier; see website for detailed description www.perinatreg.nl.13

The board of The Netherlands Perinatal Registry granted permission to use the anonymous

registry data for this study. The Netherlands Perinatal Registry has been extensively

described and used in several recent studies.1,2,4,10,13-15

Assignment of high risk

In Dutch obstetric care a pregnancy is considered a valid high risk, justifying referral, if an

adverse perinatal outcome, adverse maternal outcome or combination of both is present

or is to be expected. Indications for referral are listed in the so-called ‘List of Obstetric

Indications’ (in Dutch: Verloskundige Indicatie Lijst).5 Community midwives are trained in

the use of the ‘List of Obstetric Indications’ to detect (expected) high risks.

Judgment of adequacy of high and low risk assignment

As a retrospective measure to judge whether assumed low risk women truly were low risk,

we used the prevalence of so-called ‘Big4’ morbidities as an indicator of risk status (gold

standard).2,15 From a detailed analysis of The Netherlands Perinatal Registry we know that

four specific morbidities precede perinatal mortality in 85% of cases, the so-called ‘Big4’

morbidities.2,15 These are: congenital anomalies (list defined), preterm birth (<37th week

of gestation), small for gestational age (SGA, birthweight <10th percentile for gestational

age16) or low Apgar score (<7, 5 minutes after birth). Congenital anomalies are registered

postpartum through a standard coding system with eight different organ systems, and

further distinction into 51 specific and 20 more global categories. By using remnant ‘Big4’

morbidity among assumed low risk women as yard stick we focus on undetected risks which

are relevant to perinatal mortality. This focus by definition does not include any unexpected

adverse maternal outcome in low risk women.

Eff ectiveness of risk selection

The primary outcome in quantifying the effectiveness of risk selection is the ‘Big4’ (high

risk) prevalence at the end of pregnancy in primary care. In the theoretical perfect case

‘Big4’ morbidity is absent in assumed low risk pregnancies.

From this starting point, we utilise three methods to quantify effectiveness of risk selection:

Method 1. with a flow chart approach describing the proportional shift of women from

primary care to secondary/tertiary care over the course of pregnancy, distinguishing

115

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

between two referral moments (during pregnancy and during parturition); risk selection

should preferably take place during pregnancy;

Method 2. by comparing eight, mutually exclusive, patient subgroups in terms of the level

of care at first antenatal booking (primary, secondary/tertiary care) and subsequent referral,

where ‘Big4’ prevalence and perinatal mortality are observed in the various subgroups;

preferably, perinatal mortality is only increased in the secondary/tertiary care group with

referral of ‘Big4’ pregnancies during parturition being a rare event.

The eight subgroups were defined according to parity (primiparous/multiparous), ethnicity

(Western/non-Western), and living in a deprived neighbourhood (yes/no, based on 4-digit

zip codes and an official public list of 40 deprived zip code based neighbourhoods).17 The

eight groups presumably differ according to ‘Big4’ prevalence and care characteristics.

Method 3. by formal analysis of diagnostic performance of selection and referral of one

‘Big4’ category, i.e., SGA. The theoretical goal is to obtain a SGA-free population in the

primary care setting, which is studied for the same eight patient subgroups as before.

If the subgroup SGA prevalence matches the subgroup variation in test characteristics,

then, the selected patient subgroup factors (parity/ethnicity/neighbourhood) are likely

to be responsible for the between subgroup differences in test characteristics. However, if

there is a discrepancy, other factors may explain the subgroup test characteristics variation

factors, e.g., system related factors.

SGA was chosen as, together with congenital anomalies, it can be detected the easiest in

the antenatal phase. Moreover, there is general consensus on (improving) detection of SGA

because of the inherent increased risk for adverse outcome18, and most congenital anomalies

are now detected by routine ultrasound (introduced in 2007) at 20 weeks of gestational age.

Referral categories

To describe the referral process we defined five mutually exclusive categories:

I. First antenatal booking in secondary/tertiary care, no referral by definition, birth in

hospital in secondary/tertiary care.

II. First antenatal booking in primary care, referral during pregnancy, birth in hospital in

secondary/tertiary care;

III. First antenatal booking in primary care, referral during parturition, birth in hospital in

secondary/tertiary care;

116

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

IV. First antenatal booking in primary care, no referral, home birth in primary care;

V. First antenatal booking in primary care, no referral, short-stay hospital birth in primary

care.

Diagnostic performance of SGA referral and selection

For the third quantification of effectiveness of risk selection we applied a diagnostic test

framework in which SGA selection (and referral) during pregnancy and during parturition

are treated as a positive ‘test result’, whereas the presence of SGA at birth is regarded as

the ‘gold standard’ outcome. The related 2x2 table is illustrated in table 7.1: ‘A’ represents

true positives (referral of SGA), ‘B’ false positives (referral of non-SGA), ‘C’ false negatives

(non-referral of SGA), and ‘D’ true negatives (non-referral of non-SGA).

The following five diagnostic test characteristics were determined applying method 319-21:

• Sensitivity [A/(A+C)], the proportion of SGA cases referred;

• Specificity [D/(B+D)], the proportion of non-SGA cases which are not referred;

• Negative predictive value (NPV) [D/(C+D)], the proportion of non-referred women

without a SGA baby;

• Negative likelihood ratio (LR-), [1-sensitivity/specificity], determines whether a negative

‘test’ result, i.e., no referral, decreases the probability of having a SGA baby for women

who are not referred. A negative likelihood ratio ranging from 0 to <1 implies diagnostic

value of the test, a value of 1 represents a test without diagnostic value (‘similar to

flipping a coin’) regarding SGA selection and referral. The lower the LR-, the higher the

diagnostic value, i.e., non-referral being associated with absence of SGA;

• SGA false negative rate (FN) [1-NPV], i.e., the proportion of non-referred women with

a SGA baby (false negative).

Table 7.1 Main 2x2 table with actual referral to secondary/tertiary care treated as a positive ‘test result’ and the presence of SGA (small for gestational age) at birth treated as the ‘gold standard’ outcome

SGA present at birth

Yes No

ReferralYes A B

No C D

A: True positive / B: False positive / C: False negative / D: True negative

117

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

The NPV, LR- and FN can be considered the most important test characteristics. These

characteristics pertain to the least desirable situation, i.e., a woman with a high risk (e.g.,

SGA) pregnancy giving birth in primary care which is only intended for low risk pregnancies.

RESULTS

Process of risk selection

A total of 1,407,387 single pregnancies were analysed. Figure 7.1 displays the selection

and referral of ‘Big4’ pregnancies by referral category in a flowchart; 15% of all pregnancies

are ‘Big4’ pregnancies. The dark grey area represents the ‘Big4’ proportion in primary care

(above the dashed line) and in secondary/tertiary care (below the dashed line). The dark

grey area above the dashed line diminishes in width if ‘Big4’ pregnancies are referred from

primary care to secondary/tertiary care during pregnancy or during parturition. Over the

course of pregnancy, the proportion of ‘Big4’ pregnancies in primary care decreases from

14% at first antenatal booking to 6% in women giving birth at home and to 9% among

women with a short-stay hospital delivery under the supervision of a community midwife.

Table 7.2 displays demographics and outcomes in the overall study population by the referral

categories. Most women are multiparous (54%), between the ages of 20-35 years (84%),

of Western origin (84%) and living in a non-deprived neighbourhood (94%). The largest

referral group is the group of women referred during pregnancy (group II, n=466,415). In

the group referred during parturition (group III), 70% of women are primiparous compared

to 48% of women referred during pregnancy (group II). Group IV has the lowest risks, the

group with the highest risk is group I.

Method 2: ‘Big4’ and perinatal mortality prevalence by patient subgroup and referral category

Table 7.3 shows the ‘Big4’ and perinatal mortality (overall 9.8 per 1,000) prevalence for the

different subgroups in the five referral categories. Late ‘Big4’ referral, i.e., ‘Big4’ prevalence

in women referred during parturition is not rare with a range of 14-19%; perinatal mortality,

however, is relatively low ranging from 3.5 to 8.3 per 1,000 births.

Overall, there are large differences in perinatal mortality and ‘Big4’ prevalence between

subgroups and referral categories: in all referral categories, primiparous women have

118

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

Fig

ure

7.1

Re

ferr

al a

nd ri

sk s

elec

tion

in D

utch

obs

tetr

ic c

are

in a

bsol

ute

num

bers

; effe

ctiv

enes

s of

risk

sel

ectio

n is

illu

stra

ted

as th

e pr

opor

tion

of

‘Big

4’ p

regn

anci

es in

dar

k gr

ey. T

he d

iffer

ent r

efer

ral c

ateg

orie

s ar

e lig

ht g

rey;

from

left

to ri

ght r

efer

ral c

ateg

orie

s I t

o V.

119

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

Ta

ble

7.2

D

emog

raph

ics a

nd o

utco

me

in th

e st

udy

popu

latio

n by

the

five

diffe

rent

refe

rral

cat

egor

ies

 Re

ferr

al c

ateg

orie

s 

Refe

rral

cat

egor

yI

IIIII

IVV

Tota

l

Firs

t boo

king

Seco

ndar

y/Te

rtia

ry c

are

Prim

ary

care

Prim

ary

care

Prim

ary

care

Prim

ary

care

Birt

hSe

cond

ary/

Tert

iary

car

eSe

cond

ary/

Tert

iary

car

eSe

cond

ary/

Tert

iary

car

ePr

imar

y ca

re:

hom

ePr

imar

y ca

re:

hosp

ital

Refe

rral

No

refe

rral

Dur

ing

preg

nanc

yD

urin

g pa

rtur

ition

No

refe

rral

No

refe

rral

N25

3,86

746

6,41

519

0,34

533

6,28

216

0,47

81,

407,

387

Parit

y*Pr

imip

arou

s (%

)12

1,06

6 (4

8)22

6,13

3 (4

8)13

4,10

2 (7

0)10

44,8

8 (3

1)63

,567

(40)

649,

356

(46)

Mul

tipar

ous

(%)

132,

794

(52)

240,

121

(52)

56,1

93 (3

0)23

1,71

9 (6

9)96

,826

(60)

757,

653

(54)

Mat

erna

l age

*<2

0 ye

ars

(%)

4,72

5 (2

)**

7,53

9 (2

)4,

333

(2)

2,82

0 (1

)4,

637

(3)

24,0

54 (2

)

20-3

5 ye

ars

(%)

206,

984

(82)

**38

2,19

5 (8

2)16

6,93

8 (8

8)29

0,44

2 (8

6)13

6,41

8 (8

5)1,

182,

977(

84)

>35

year

s (%

)42

,158

(17)

**76

,681

(16)

19,0

74 (1

0)43

,020

(13)

19,4

23 (1

2)20

0,35

6 (1

4)

Ethn

icity

Wes

tern

(%)

214,

332

(84)

387,

985

(83)

157,

245

(83)

308,

425

(92)

112,

775

(70)

1,18

0,76

2 (8

4)

Non

-Wes

tern

(%)

39,5

35 (1

6)78

,430

(17)

33,1

00 (1

7)27

,857

(8)

47,7

03 (3

0)22

6,62

5 (1

6)

Tabl

e 7.

2 co

ntin

ues o

n ne

xt p

age.

120

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

Ta

ble

7.2

Co

ntin

ued

 Re

ferr

al c

ateg

orie

s 

Refe

rral

cat

egor

yI

IIIII

IVV

Tota

l

Firs

t boo

king

Seco

ndar

y/Te

rtia

ry c

are

Prim

ary

care

Prim

ary

care

Prim

ary

care

Prim

ary

care

Birt

hSe

cond

ary/

Tert

iary

car

eSe

cond

ary/

Tert

iary

car

eSe

cond

ary/

Tert

iary

car

ePr

imar

y ca

re:

hom

ePr

imar

y ca

re:

hosp

ital

Refe

rral

No

refe

rral

Dur

ing

preg

nanc

yD

urin

g pa

rtur

ition

No

refe

rral

No

refe

rral

Nei

ghbo

rhoo

dN

on-d

epriv

ed23

7,83

6 (9

4)43

7,35

8 (9

4)17

7,65

2 (9

3)32

4,47

7 (9

6)14

3,69

9 (9

0)1,

321,

022

(94)

Dep

rived

16,0

31 (6

)29

,057

(6)

12,6

93 (7

)11

,805

(4)

16,7

79 (1

0)86

,365

(6)

Cong

enita

l ano

mal

ies

No

(%)

245,

554

(97)

447,

192

(96)

185,

515

(97)

332,

647

(99)

158,

329

(99)

1,36

9,23

7 (9

7)

Yes

(%)

8,31

3 (3

)19

,223

(4)

4,83

0 (3

)3,

635

(1)

2,14

9 (1

)38

,150

(3)

Pret

erm

birt

hN

o (%

)22

3,94

6 (8

8)42

6,32

4 (9

1)17

7,40

4 (9

3)33

4,92

0 (1

00)

157,

945

(98)

1,32

0,53

9 (9

4)

Yes

(%)

29,9

21 (1

2)40

,091

(9)

12,9

41 (7

)1,

362

(0)

2,53

3 (2

)86

,848

(6)

Smal

l for

ges

tatio

nal a

geN

o (%

)23

1,49

4 (9

1)42

5,52

2 (9

1)17

8,06

2 (9

4)32

1,10

7 (9

5)15

0,53

2 (9

4)1,

306,

717

(93)

Yes

(%)

22,3

73 (9

)40

,893

(9)

12,2

83 (6

)15

,175

(5)

9,94

6 (6

)10

0,67

0 (7

)

* Tot

als

do n

ot a

lway

s ad

d up

due

to m

issi

ng v

alue

s (p

arity

378

mis

sing

s, m

ater

nal a

ge 2

84 m

issi

ngs)

.**

Per

cent

ages

add

up

to 1

01%

bec

ause

of r

ound

ing

121

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

Ta

ble

7.3

Pr

eval

ence

of ‘

Big4

’ mor

bidi

ties

and

perin

atal

mor

talit

y by

the

five

refe

rral

cat

egor

ies

in d

iffer

ent s

ubgr

oups

(bas

ed o

n pa

rity,

eth

nici

ty

and

neig

hbou

rhoo

d). T

otal

num

ber o

f pre

gnan

cies

(N=1

,407

,009

) doe

s no

t inc

lude

378

par

ity m

issi

ngs.

  

Refe

rral

cat

egor

ies

  

 

Refe

rral

cat

egor

yI

 II

 III

 IV

 V

 To

tal

Firs

t boo

king

Seco

ndar

y/Te

rtia

ry

care

Prim

ary

care

Prim

ary

care

Prim

ary

care

Prim

ary

care

Birt

hSe

cond

ary/

Tert

iary

ca

reSe

cond

ary/

Tert

iary

ca

reSe

cond

ary/

Tert

iary

ca

rePr

imar

y ca

re: h

ome

Prim

ary c

are:

hos

pita

l

Refe

rral

No

refe

rral

 D

urin

g pr

egna

ncy

 D

urin

g pa

rtur

ition

 N

o re

ferr

al 

No

refe

rral

 

  

NBi

g4*

†**

NBi

g4*

†**

NBi

g4*

†**

NBi

g4*

†**

NBi

g4*

†**

NBi

g4*

†**

Su

bg

rou

ps

Prim

ipar

ous

Wes

tern

Non

-dep

rived

101,

876

25%

20.3

188,

323

24%

13.8

110,

669

16%

3.7

94,4

359%

0.8

44,8

8212

%5.

954

0,18

519

%10

.0

Dep

rived

3,28

428

%26

.56,

379

27%

16.3

3,71

918

%3.

52,

597

8%0.

82,

063

15%

3.4

18,0

4221

%11

.8

Non

-Wes

tern

Non

-dep

rived

12,2

9330

%34

.224

,570

27%

17.6

15,3

1618

%5.

35,

947

12%

1.7

12,3

0214

%5.

170

,428

22%

14.3

Dep

rived

3,61

330

%39

.66,

861

29%

18.4

4,39

819

%5.

51,

509

13%

2.7

4,32

014

%2.

820

,701

23%

14.9

Mul

tipar

ous

Wes

tern

Non

-dep

rived

106,

312

18%

19.6

188,

639

15%

11.7

41,7

2514

%6.

220

8,15

05%

0.6

63,8

006%

3.6

608,

626

11%

8.1

Dep

rived

2,85

827

%33

.94,

522

20%

13.0

1,09

716

%7.

33,

179

6%1.

91,

993

8%2.

013

,649

16%

12.7

Non

-Wes

tern

Non

-dep

rived

17,3

4823

%31

.535

,683

18%

15.6

9,89

715

%7.

015

,876

7%0.

622

,640

7%2.

510

1,44

415

%12

.2

Dep

rived

6,27

625

%33

.511

,277

20%

17.0

3,47

415

%8.

34,

514

6%0.

98,

393

8%1.

433

,934

15%

13.2

Tota

l25

3,86

022

%22

.346

6,25

420

%13

.519

0,29

516

%4.

733

6,20

76%

0.7

160,

393

9%4.

01,

407,

009

15%

9.8

* ‘Bi

g4’ m

orbi

ditie

s in

per

cent

age

of th

e to

tal a

mou

nt o

f pre

gnan

cies

in th

at s

ubgr

oup.

** P

erin

atal

mor

talit

y (fr

om 2

2 w

eeks

of g

esta

tiona

l age

unt

il 7

days

pos

tpar

tum

) per

1,0

00 b

irths

.

122

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

higher ‘Big4’ prevalences compared to multiparous women (range 8-30% versus 5-27% for

multiparous women). Perinatal mortality, however, is lower in primiparous women referred

during parturition compared to multiparous women (3.5-5.5 vs. 6.2-8.3 per 1,000 births,

respectively). In almost all subgroups and referral categories, non-Western women have

higher ‘Big4’ and perinatal mortality prevalences; outcome differences between living in a

non-deprived versus living in a deprived neighbourhood are smaller.

Method 3: SGA selection during pregnancy

Table 7.4 shows the SGA prevalence before and after selection (and subsequent referral),

and diagnostic test characteristics for SGA selection and referral during pregnancy and during

parturition. Of all women exposed to the selection and referral process during pregnancy

(referral categories II to V), non-Western primiparous women in deprived neighbourhoods

have the highest SGA prevalence (13%). For all subgroups, SGA prevalence is lower after

selection compared to before. Sensitivity and specificity of SGA selection during pregnancy

ranges from 49% to 59% and from 58% to 63% respectively. The range of NPV is 89-97%.

The negative likelihood ratio (LR-) ranges from 0.69 to 0.85, indicating that that the SGA

prevalence decreases (after the selection / ’test’) in the group of women who are not referred

during pregnancy. However, the values are close to 1 (which would imply a test without

diagnostic value) which implies a modest discriminating value. The FN ranges from 3-11%

depicting the percentage of non-referred women having SGA babies.

Method 3: SGA selection during parturition

Of all women exposed to the selection and referral process during parturition (referral

categories III to V) primiparous non-Western women have the highest SGA prevalence

(11%). For all subgroups, there is no difference in SGA prevalence before and after selection.

Sensitivity and specificity of SGA selection during parturition range from 15% to 45% and

from 54% to 87%, respectively. For primiparous women the sensitivity is higher than for

multiparous women. Specificity, on the other hand, is higher for multiparous women than

for primiparous women. The NPV is similar to that for the SGA selection process during

pregnancy 89-97%. The LR- ranges from 0.97 to 1.05 implying no diagnostic value of risk

selection for SGA during parturition. The FN ranges from 3-11%.

123

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

Ta

ble

7.4

Su

bgro

up S

GA

pre

vale

nce

befo

re (‘

SGA

% B

F’) a

nd a

fter

(‘SG

A%

AF’

) sel

ectio

n, a

nd d

iagn

ostic

test

cha

ract

eris

tics*

; bot

h fo

r sel

ectio

n an

d re

ferr

al o

f SG

A p

regn

anci

es d

urin

g pr

egna

ncy

and

durin

g pa

rtur

ition

Sele

ctio

n du

ring

preg

nanc

ySe

lect

ion

durin

g pa

rtur

ition

SGA

%

BFSG

A%

A

FSE

NS

SPEC

NPV

LR-

FNSG

A%

BF

SGA

%

AF

SEN

SSP

ECN

PVLR

-FN

Su

bg

rou

ps

Prim

ipar

ous

Wes

tern

Non

-dep

rived

9%7%

53%

58%

93%

0.82

7%7%

7%42

%56

%93

%1.

057%

Dep

rived

11%

9%53

%58

%91

%0.

809%

9%9%

44%

56%

91%

1.01

9%N

on-W

este

rnN

on-d

epriv

ed12

%11

%50

%59

%89

%0.

8511

%11

%11

%45

%54

%89

%1.

0111

%D

epriv

ed13

%11

%49

%61

%89

%0.

8311

%11

%11

%44

%57

%89

%0.

9911

%

Mul

tipar

ous

Wes

tern

Non

-dep

rived

4%3%

52%

63%

97%

0.76

3%3%

3%15

%87

%97

%0.

983%

Dep

rived

7%5%

59%

59%

95%

0.69

5%5%

5%18

%83

%95

%0.

995%

Non

-Wes

tern

Non

-dep

rived

6%5%

53%

58%

95%

0.81

5%5%

5%23

%80

%95

%0.

975%

Dep

rived

7%5%

52%

60%

95%

0.79

5%5%

5%21

%79

%95

%1.

005%

Tota

l7%

5%52

%60

%95

%0.

795%

5%5%

33%

73%

95%

0.93

5%

* SE

NS:

sen

sitiv

ity, S

PEC:

spe

cifi c

ity, N

PV: n

egat

ive

pred

ictiv

e va

lue,

LR-

: neg

ativ

e lik

elih

ood

ratio

, FN

: SG

A fa

lse

nega

tive

rate

.

124

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

DISCUSSION

Principal fi ndings

To our knowledge, this is the largest study on the effectiveness of risk selection in Dutch

primary obstetric care. The main focus was to examine whether risk selection realises its

aim of a true low risk group of pregnant women by identifying and referring high risk

pregnancies. Even though many ‘Big4’ pregnancies are referred, our results demonstrate that

a true low risk population is never attained, with a ‘Big4’ prevalence of up to 15% in primary

care (table 7.3), intended for low risk pregnancies only. Also, ‘Big4’ prevalence among late

referrals (during parturition) was still substantial, ranging from 14% to 19%.

We further observed a suboptimal discrimination of SGA and non-SGA pregnancies (low

sensitivity, LR- close to 1 in all subgroups) with a SGA prevalence of 3% to 11% still being

born in primary care.

Moreover, we observed a discrepancy in the subgroup SGA prevalence and the subgroup

variation in SGA selection test characteristics (table 7.4). This implies other factors to be

responsible for the suboptimal test characteristics instead of the selected patient factors,

e.g., system related factors. One may think of differences in availability of SGA screening

methods.

Home birth versus short-stay hospital birth

In primary care, short-stay hospital births showed higher ‘Big4’ prevalence compared to

home births (9% vs. 6%). This may reflect an unintentional selection process by either the

midwife or self-selection by pregnant women, i.e., more healthy women appear to opt for

home birth. This has also been observed in other studies.22

Preventability

From our results, the question arises whether the birth of a ‘Big4’ baby in a primary rather

than secondary care setting is a preventable situation. As stated, congenital anomalies

and SGA are better predictable than (spontaneous) preterm birth and a low Apgar score.

In accepting that a NPV of 100% is not attainable, we actually state that ‘Big4’ deliveries in

a primary care are to some extent inevitable in a system with different risk-based settings.

125

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

The subsequent question then refers to the observed ‘setting safety’ of a primary care

setting. Both general consensus and the ‘List of Obstetric Interventions’ agree that a neonate

with a ‘Big4’ morbidity is better off in a hospital setting under the care of an obstetrician/

paediatrician.23 The benefit of this so-called ‘setting safety’ may be related to availability of

neonatological expertise, continuous fetal heart rate monitoring or advanced resuscitation

equipment.4,10

As we showed less optimal SGA selection, and more referrals during parturition for

primiparous women, we believe that all primiparous women should deliver in a hospital

environment, either under supervision of a midwife (birth centre) or an obstetrician. This

also follows from unequivocal evidence from previous studies24-26, where others generally

waive the primary care option.24

Strengths and limitations

The strengths of the current study include the nationwide approach and high coverage of

pregnancies in The Netherlands Perinatal Registry over a long period of time. In addition,

the use of ‘Big4’ morbidities, as the major precursors of perinatal mortality, allows for an

easy-to-comprehend proxy measure of high risk pregnancy.2,15 Also, the application of a

diagnostic test framework allows the results to be interpreted objectively in a standardised

way; comparisons with other diagnostic test studies can be easily made. Another strength

pertains to the use of subgroups. It is interesting at the very least to see a discrepancy

between the level of subgroup differences for SGA prevalence compared to the smaller

subgroup differences in test characteristics.

This study also has limitations. Firstly, it is not possible to determine the exact indication for

why women were referred because of the retrospective nature of The Netherlands Perinatal

Registry. For our study objective, the effect of this limitation is limited as we focused on high

risk births taking place in primary care, which is intended for low risk births. Our estimate

of prevalence of high risk births in primary care is conservative as it is likely to be higher,

providing it would have been possible to take into account all referral indications. Another

possible limitation is that with the ‘Big4’ approach, maternal and other non-‘Big4’ related

risks are disregarded. This problem also appears to be limited as the majority of referral

indications according to the ‘List of Obstetric Interventions’ pertain to fetal/neonatal risks

alone.5 Finally, the impact of routine ultrasound examination at 20 weeks of gestational age

(introduced in 2007) on congenital abnormality rates and perinatal mortality rates due to

second trimester abortions cannot be evaluated in our 2000-2007 dataset.

126

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

Previous studies on risk selection in the Dutch system

Our findings contradict most previous studies on the effectiveness of risk selection in

Dutch primary obstetric care, stating that risk selection is effective.6-9 In contrast with our

study, these studies took into account maternal morbidity and obstetric interventions

(e.g., caesarean section). However, these studies have been conducted some time ago, are

restricted to smaller study groups, specific regions or did not evaluate the risk selection

process systematically.6-9 Another limitation of previous studies is that they defined the

effectiveness of risk selection not only as prevention of adverse perinatal outcomes

but also as prevention of obstetric interventions such as a caesarean section.7-9 While

we recognise that the assessment of risk selection must be weighed against the risk of

possibly unnecessary obstetric interventions, the primary goal of adequate risk selection

and subsequent referral is to prevent adverse perinatal and/or maternal outcomes, the

prevention of (unnecessary) obstetric interventions being an important secondary goal.6,9,27

Several reports have expressed concern on the effectiveness of risk selection in Dutch

primary obstetric care.2,4,10 A recent Dutch study revealed that in 43% of Neonatal Intensive

Care Unit (NICU) admissions the pregnancy had been indicated as low risk, and thus

parturition had started in primary obstetric care.10 Furthermore, infants of pregnant women

at supposedly low risk whose labor started in primary care had a significantly higher delivery

related perinatal mortality risk than the infants of assumed high risk women whose labor

started in secondary care (relative risk 2.33, confidence interval 1.12-4.83).4 Infants of women

who were referred during parturition had a 3.66 times higher risk of delivery related perinatal

mortality than infants of women who started labor in secondary care, and a 2.5-fold higher

risk of NICU admission.4 These studies emphasise that the level of healthcare provision

could be improved for a proportion of supposedly low risk pregnant women at the onset

of labor. Whether the delay in referral is related to late diagnosis (no continuous fetal heart

rate monitoring during parturition in primary care), transport to hospital or assessment

(‘primary care is supposedly low risk’), is yet unclear.4,10

Possible implications

Our results demonstrate that the aim of risk selection in Dutch primary obstetric care

is suboptimally attained. We propose some directions of improvement. As stated in the

‘List of Obstetric Interventions’, risk selection is currently exclusively done by primary

care community midwives. Possible improvements could be the increase of midwives’

competence and capabilities, or introduction of a checklist-based standardised risk

127

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

selection strategy at first antenatal booking. However, we believe that the required pace

of change is more likely to be achieved through a combination of the latter with ‘shared

care’: better cooperation between midwives and obstetricians who are jointly responsible

for the determination of a woman’s risk status, thereby joining their expertise which is either

physiology-based (midwives) or pathology-based (obstetricians).28 Shared obstetric care

has already been implemented in some form in other Western countries such as Australia

and the United Kingdom.29-31 One study demonstrated a 27% increase in the detection rate

of intrauterine growth restriction for women receiving shared obstetric care as opposed to

conventional obstetric care.32 For more generalisable results however, a study to evaluate

different shared care strategies has to be conducted in The Netherlands because of the

unique system of obstetric care. We believe that our recommendation for shared care also

applies to countries which are considering or already have an obstetric care system with

features similar to the Dutch system, such as Canada.33-36

REFERENCES1. Mohangoo AD, Buitendijk SE, Hukkelhoven CW, et

al. Higher perinatal mortality in The Netherlands than in other European countries: the Peristat-II study. Ned Tijdschr Geneeskd. 2008;152:2718-2727.

2. Bonsel GJ, Birnie E, Denktas S, Poeran J, Steegers EAP. Dutch report: Lijnen in de Perinatale Sterfte, Signalementstudie Zwangerschap en Geboorte 2010. Rotterdam: Erasmus MC; 2010.

3. Sheldon T. Obstetric care must change if Nether-lands is to regain reputation for safe childbirth. BMJ. 2008;336:c239.

4. Evers AC, Brouwers HA, Hukkelhoven CW, et al. Perinatal mortality and severe morbidity in low and high risk term pregnancies in the Netherlands: prospective cohort study. BMJ. 2010;341:c5639.

5. Amelink-Verburg MP, Buitendijk SE. Pregnancy and labour in the Dutch maternity care system: what is normal? The role division between mid-wives and obstetricians. J Midwifery Womens Health. 2010;55:216-225.

6. Bais JMJ, Pel M. The basis of the Dutch obstetric system: risk selection. Eur Clinics Obstet Gynaecol. 2007;2:209-212.

7. Eskes M. Het Wormerveeronderzoek. Meer-jarenonderzoek naar de kwaliteit van de verlos-kundige zorg rond een vroedvrouwenpraktijk. Thesis 1989, University of Amsterdam.

8. Berghs G, Spanjaards E. De normale zwanger-schap: bevalling en beleid. Thesis 1988, Catholic University Nijmegen.

9. Bais JMJ. Risk selection and detection. A critical appraisal of the Dutch obstetric system. Thesis 2004, University of Amsterdam.

10. Evers AC, van Leeuwen J, Kwee A, et al. Mortality and morbidity among full-term neonates in a neonatal intensive care unit in the Utrecht region, the Netherlands. Ned Tijdschr Geneeskd. 2010;154:118.

11. Sheldon T. Prompt access to expert care is among advice to reduce perinatal deaths in Netherlands. BMJ. 2010;340:c277.

12. World Health Organization. Neonatal and perinatal mortality. Country, regional and global estimates. Geneva: WHO, 2006. (online: http://whqlibdoc.who.int/publications/2006/9241563206_eng.pdf).

13. Perinatal care in The Netherlands 2007 (in Dutch: Perinatale zorg in Nederland 2007). Utrecht: The Netherlands Perinatal Registry; 2009.

128

RISK SELECTION

IN D

UTCH

OBSTETRIC CA

RE

7

14. Tromp M, Meray N, Ravelli AC, Reitsma JB, Bonsel GJ. Medical Record Linkage of Anonymous Registries without Validated Sample Linkage of the Dutch Perinatal Registries. Stud Health Technol Inform. 2005;116:125-130.

15. Van der Kooy J, Poeran J, de Graaf JP, et al. Planned home compared with planned hospital births in the Netherlands: intrapartum and early neonatal death in low-risk pregnancies. Obstet Gynecol. 2011;118:1037-1046.

16. Kloosterman GJ. On intrauterine growth. Int J Gynaecol Obstet. 1970;8:895-912.

17. Gent, van WPC. Realistic regeneration -housing contexts and social outcomes of neighbourhood interventions in Western European cities-; chapter 4. Thesis 2009, University of Amsterdam. (online: http://dare.uva.nl/document/153004).

18. Figueras F, Gardosi J. Intrauterine growth restric-tion: new concepts in antenatal surveillance, di-agnosis, and management. Am J Obstet Gynecol. Apr 2011;204:288-300.

19. Altman DG, Bland JM. Diagnostic tests 2: Predictive values. BMJ. 1994;309:102.

20. Altman DG, Bland JM. Diagnostic tests. 1: Sensitivity and specificity. BMJ. 1994;308:1552.

21. Deeks JJ, Altman DG. Diagnostic tests 4: likelihood ratios. BMJ. 2004;329:168-169.

22. Kennare RM, Keirse MJ, Tucker GR, Chan AC. Planned home and hospital births in South Aus-tralia, 1991-2006: differences in outcomes. Med J Aust. 2010;192:76-80.

23. National Institute for Health and Clinical Excellence (see Antenatal Care: Routine Care for the Healthy Pregnant Woman). (online: http://www.nice.org.uk/nicemedia/live/11947/40115/40115.pdf).

24. Chervenak FA, McCullough LB, Brent RL, Levene MI, Arabin B. Planned home birth: the professional responsibility response. Am J Obstet Gynecol. 2013;208:31-38.

25. Brocklehurst P, Hardy P, Hollowell J, et al. Perinatal and maternal outcomes by planned place of birth for healthy women with low risk pregnancies: the Birthplace in England national prospective cohort study. BMJ. 2011;343:d7400.

26. De Neef T, Hukkelhoven CW, Franx A, van Everhardt E. Uit de lijn der verwachting. Nederl Tijdschrift Obstet Gynaecol 2009;122:34-342.

27. National Institute for Health and Clinical Excel-lence. Intrapartum care: care of healthy women and their babies during childbirth. CG55. 2007. (online: www.nice.org.uk/nicemedia/pdf/IPC-NICEGuidance.pdf ).

28. Posthumus AG, Scholmerich VL, Waelput AJ, et al. Bridging Between Professionals in Perinatal Care: Towards Shared Care in The Netherlands. Matern Child Health J. 2012 (in press).

29. Halliday J. WUDWAW ‘Who Usually Delivers Whom and Where’, Perinatal Data Collection Unit. Victoria, Australia: Victorian Government Department of Human Services; 1999.

30. Hampson JP, Roberts RI, Morgan DA. Shared care: a review of the literature. Fam Pract. 1996;13:264-279.

31. Tucker J, Florey CD, Howie P, McIlwaine G, Hall M. Is antenatal care apportioned according to obstetric risk? The Scottish antenatal care study. J Public Health Med. 1994;16:60-70.

32. Chan FY, Pun TC, Tse LY, Lai P, Ma HK. Shared antenatal care between family health services and hospital (consultant) services for low risk women. Asia Oceania J Obstet Gynaecol. 1993;19:291-298.

33. De Vries R, Benoit C, Van Teijlingen E, Wrede S. Birth by design: pregnancy, maternity care and midwifery in North America and Northern Europe. New York: Routledge, 2001.

34. Hatem M, Sandall J, Devane D, Soltani H, Gates S. Midwife-led versus other models of care for childbearing women. Cochrane Database Syst Rev. 2008:CD004667.

35. Bick D. Enhancing safety in the maternity services: a greater role for midwife-led care? Midwifery. 2009;25:1-2.

36. Royal College of Obstetricians and Gynaecologists/Royal College of Midwives. Home births. Joint statement No.2; 2007 (online: http://www.rcog.org.uk/files/rcog-corp/uploaded-files/JointStatmentHomeBirths2007.pdf).

8

Obstetrics and Gynecology. 2011;118:1037-46.

Planned home compared with planned hospital births in The Netherlands: intrapartum and

early neonatal death in low-risk pregnancies

J. van der Kooy, J. Poeran, J.P. de Graaf, E. Birnie, S. Denktaş, E.A.P. Steegers, G.J. Bonsel

130

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

ABSTRACTObjective The purpose of our study was to compare the intrapartum and early neonatal

mortality rate of planned home birth vs. planned hospital birth in community midwife-led

deliveries, after case mix adjustment.

Methods Perinatal outcome of 679,952 low-risk women was obtained from the Dutch

Perinatal Registry (2000-2007). This group represents all women who had a choice between

home and hospital birth. Two different analyses were performed; natural prospective

approach (intention-to-treat like analysis) and perfect guideline approach (per-protocol like

analysis). Unadjusted and adjusted odds ratios were calculated. Case mix was based on the

presence of at least one of the following: congenital abnormalities, small for gestational age,

preterm birth, or low Apgar score. We also investigated the potential risk role of intended

place of birth. The technique used was multivariable stepwise logistic regression.

Results Intrapartum and neonatal death 0-7 days was observed in 0.15% of planned

home vs 0.18% in planned hospital births (crude RR 0.80 95%CI 0.71-0.91). After case mix

adjustment, the relation is reversed, showing non-significant increased mortality risk of

home birth (OR 1.05 95%CI 0.91-1.21). In certain subgroups additional mortality may arise

at home if risk conditions emerge at birth (up to 20% increase).

Conclusion Home birth, under routine conditions, is generally not associated with

increased intrapartum and early neonatal death, yet in subgroups additional risk cannot

be excluded.

131

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

INTRODUCTIONThe debate on the safety of home births continues in the literature as recently addressed

in the Lancet.1 In the Netherlands, approximately 50% of women give birth under the

supervision of a community midwife. The community midwives are independent health

care professionals in the Netherlands, operating either solely or in group practices. The

proportion of home birth deliveries in the Netherlands has steadily decreased over the

last decade but is currently stable at 25% of all births. Several Anglo-Saxon countries are

considering the reintroduction of home births, based on recent claims of sufficient safety.2

The reverse trend is observed in the Netherlands, where the debate has intensified since

the national perinatal mortality rate showed to be one of the highest in Europe.3

In the Dutch system, independently operating community midwives provide care for low-

and medium-risk pregnant women (primary healthcare). High-risk pregnant women are

referred to the gynaecologist for remaining ante- and intrapartum care. If no or only a few

risk factors are present, women can stay with the midwife and decide where the delivery

will take place: at home or in the hospital, both supervised by the community midwife.

For pregnant women with so called ‘medium-risk’ delivery in hospital is obligatory but can

still be under the supervision of the community midwife. A strict definition of medium

risk, created and agreed upon by midwives and gynaecologists together, is defined in the

Dutch guidelines.4 The claimed benefits of planned home births include the reduction of

maternal-fetal morbidity, a lower risk for unjustified medical interventions, and psychosocial

advantages for the mother. These benefits may be counterbalanced by the disadvantages

associated with a high intrapartum referral rate and an increased perinatal mortality,

morbidity and long term negative effects.5-11

This paper re-addresses the Dutch evidence focusing on two critical features of previous

analyses. First, previous studies compared outcomes after exclusion of pregnant women

who in view of the delivery guidelines should have been referred to a gynaecologist. Second,

previous studies did not apply case mix analysis, assuming risk equivalence of home and

hospital groups.5,9,12-18 Case mix may, however, differ across planned place of delivery, due

to self selection or due to the midwife’s proposal, with the healthiest and most affluent

women receiving home birth (confounding the comparison by indication bias).5 6,7,11,19-21

The purpose of our study was to compare the intrapartum and early neonatal mortality

rate of planned home birth vs. planned hospital birth in community midwife-led deliveries,

after case mix adjustment. We compared a natural prospective approach without ex post

exclusion of unsuitable midwife cases (intention-to-treat like), with the conventional

132

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

approach based on a theoretical midwife population under perfect guideline adherence

(per-protocol like). We hypothesised that while in general no difference may exist between

home and hospital outcomes, for specific risk groups the hospital setting is protective as

obstetrical and neonatal expertise and clinical facilities are directly available (so-called

“setting safety”).

METHODS

Data

The Netherlands Perinatal Registry (PRN) contains population-based information of 96%

of all pregnancies in The Netherlands. Source data are collected by 95% of midwives,

99% of gynaecologists and 68% of paediatricians (including 100% of Neonatal Intensive

Care Unit paediatricians).3,22 (See website for detailed description: www.perinatreg.nl). We

selected the records of all singleton pregnant women, under supervision of a community

midwife at the onset of labour between 2000-2007 (693,592 women). The onset of labour

was defined as spontaneous contractions or the spontaneous rupture of membranes by

the PRN. Two subsets of pregnant women were further excluded from the original set of

693,592 women. First, 13,384 women with so called ‘medium risk’, for example women with

a history of postpartum haemorrhage or obesity (BMI>30). Dutch guidelines prescribe a

hospital delivery for these women which may be supervised by the community midwife.

Secondly we excluded records were the data was incomplete (n=256).

The remaining women (n=679,952) were categorised according to intended place of birth,

which usually is concordant with the observed place of birth either home or hospital. For

some women the place was not decided until the onset of labour. This could be due to

indifference on the part of the woman; or delayed antepartum care. The intended place

was then coded ‘unknown’. This yielded 3 intention groups: home, hospital, and, unknown.

Outcome measures, maternal and neonatal risk factors

Outcome was defined as intrapartum and early neonatal mortality , i.e. (I) intrapartum

death, (II) neonatal death up to 24 hrs, and (III) neonatal death up from 1 day to 7 days post

partum. In our low risk group under midwife supervision, mortality beyond 8 days is rare,

and regarded to be unrelated to place of delivery. The PRN does not include long term child

outcomes for example psychomotor development and behavioural function.

133

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Maternal risk factors were parity (nulliparous vs. multiparous), age, ethnicity (Western/non-

Western; based on a more refined classification in the registry), and living in a deprived

neighbourhood (yes/no, based on 4-digit zip-codes and a public list of deprived, zip-code

based, neighbourhoods issued by the Dutch government).

Detailed risk information is unavailable in national registries. Case mix of any defined

group of women was primarily represented by the prevalence (single or combined) of

‘Big4’ conditions (see below). From detailed analysis of the complete perinatal dataset of

the same Netherlands Perinatal Registry (PRN), years 2000-2007, (1.25 million records)23, it

appeared that the presence of any of 4 conditions preceded perinatal mortality in 85% of

cases. These conditions were defined as; congenital abnormalities (list defined), small for

gestational age (SGA, birth weight below the 10th percentile for gestational age, gender and

parity specific), preterm birth (< 37th week of gestation) or low Apgar score (< 7, measured

5 minutes after birth). We will continue to refer to these 4 conditions as the ‘Big4’. The main

results of this detailed analysis are found in figure 8.1.

In our current analysis these so called ‘Big4’ represent an objective estimate of the risk

challenge at birth. The preventability of their occurrence, either antenatally or during

delivery, is not at issue. Here we intentionally use it as a risk indicator, an explanatory

factor at onset. By doing so, we ignore differential management effects of setting on the

emergence of these Big4, in particular low Apgar, should they exist.

Data analysis

As primary analysis we present the results of the natural prospective approach (NPA), resem-

bling an intention-to-treat analysis. For comparison we added a perfect guideline approach

(PGA), resembling a per-protocol analysis. The NPA approach establishes, within observational

constraints, the intrapartum and early neonatal death of planned home versus planned hospital

births. It stems from the viewpoint of a pregnant woman starting birth under supervision of

a midwife (the denominator is n=679,952). The natural approach thus includes spontaneous

preterm labour since to some extent this group was not referred to the gynaecologist during

labour or was referred late during (home) delivery. Therefore a direct setting effect (admission

to hospital at the onset of labour) may be visible to the advantage of the hospital. Furthermore

indirect setting effects may be present, for example the timing of referral.

PGA includes the subset of women within the NPA population, who in retrospect were

compliant with the guidelines which define low risk at the onset of labour and therefore

allowed to choose between a home or hospital birth under supervision of a midwife.

134

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Non-compliance exists if a high risk condition was already detectable at the onset of labour.

These conditions applied to women with a gestational age <37 or >41 weeks, prolonged

rupture of membranes (>24hr) and intrauterine death with unclear timing relative to onset

of labour (see figure 8.2). PGA (n=602,331) still included undetected SGA and congenital

malformations that emerge at birth, as detection failure cannot be regarded as non-

compliance from the viewpoint of current guidelines.

First we compared characteristics of the NPA and PGA populations by intended place of

birth (t-tests for comparisons). Then we investigated the potential risk role of intended place

Figure 8.1 Perinatogram illustrating in a Venn diagram the relationship between (combinations of ) Big 4 morbidities and perinatal mortality defined as death from 22 weeks of gestational age until 7 days postpartum. In 85% of all cases of perinatal mortality, one or more Big4 morbidities are present; for instance, a low Apgar score combined with preterm birth occurs in 30.3% of all cases of perinatal mortality. *Prevalence per 1,000 births of separate and combined Big 4 morbidities and their contribution to all cases of perinatal mortality (†percentage); this adds up to 85% of all cases of perinatal mortality. The dashed circles connect low Apgar score with preterm birth and congenital abnormality with intrauterine growth restriction.

135

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

of birth by a set of predefined nested multivariable logistic regression models (stepwise

analysis; inclusion p<0.05; exclusion p>0.10) where we added maternal and neonatal (case

mix) explanat ory variables. For these variables, hospital birth was set as the reference.

All stepwise analyses were repeated with a forward and backward approach, and finally

forced inclusion of predictive variables (p<0.05). Risk factor coefficients were only shown

Figure 8.2 Flow of women through the study.

136

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

in case of significance p<0.05. Results across the three approaches were similar unless

stated otherwise.

We graphically described the crude mortality of the planned home and planned hospital

population, for the series of populations which result from successive exclusion of women

meeting a criterion for non-compliance (figure 8.3; dotted lines). This successive exclusion

through non-compliance criteria gradually transforms the NPA population into the PGA

population. If the mortality rate of a non-compliance group is average, home and hospital

mortality rates do not change on its exclusion. If the rates decrease at a different gradient

(e.g. hospital steeper than home, as after exclusion of pregnancy duration < 36 weeks) this

may point to either differential prevalence of the non-compliance factor (as here), or to

differential case fatality by setting where the largest mortality decrease is observed in the

setting with the highest case fatality (interpretable as lowest setting safety).

To support this interpretation, we first divided the crude mortality of the home and hospital

group by the respective prevalence of Big4 conditions to obtain case mix adjustment. This

assumes Big4 prevalence to be a suitable risk indicator at the group level. Subsequent

division of the resulting home/Big4 mortality ratio by the hospital/Big4 mortality yields

an index (Big4 adjusted homebirth mortality index; figure 8.3; black line). If this index is

100%, then relative mortality in home births and hospital births is equal. If the index is for

example 120%, then home births have 20% excess mortality taking our case mix differences

into account. Combining crude mortality changes with index changes allows for tentative

interpretation of setting effects.

RESULTSTable 8.1 describes the baseline characteristics of both the NPA and PGA populations

(n=679,952 vs. 602,331).

In both the NPA and PGA populations about 60% of women planned a home delivery and

about 32% planned a hospital delivery. Compared to women who planned birth in the

hospital or with unknown location, the women with planned home birth were more likely to

be multiparous, 25 years or older, of Dutch origin and to live in a privileged neighbourhood

(all of which are favourable conditions). In home birth women, neonatal case mix compared

also favourably. Premature delivery was less common, as was the prevalence of a Big4

condition (NPA home birth 8.7% vs. hospital 10.8% vs. unknown 10.5%; PGA home birth

6.5% vs. hospital 8.2% vs. unknown 7.5%; p<0.001 in both cases).

137

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Intrapartum and early neonatal mortality was 1099/679,952=1.62 ‰ in the NPA women and

551/602,331=0.91‰ in PGA women. Mortality was lower in women who were multiparous,

between 24-34 year, of Dutch origin, or living in a privileged neighbour hood (both NPA and

PGA), see table 8.1. Within the group with intrapartum and early neonatal mortality, Big4

conditions were found in 792 of the 1099 deaths (72.1%) in the NPA women, compared to

290 out of 551 deaths (52.6%) in the PGA group.

In the NPA population, crude mortality risk was significantly lower for women who planned

to give birth at home (RR 0.80 95%CI 0.71-0.91) and for women with unknown intention (RR

0.96 95%CI 0.77-1.19) compared to those who intended to give birth in hospital (P<0.05) (see

table 8.2). All maternal and neonatal risk factors, except living in a deprived neighbourhood,

showed significant effect sizes in agreement with the expected direction. Mortality was

significantly increased in infants with a Big4 outcome, especially in infants with multiple

Big4 conditions (RR 168.9 95%CI 148.9-191.4).

The nested multivariable logistic regression analysis showed that in the presence of

adjusting maternal factors only (model 2), the intended place of birth had no significant

impact on outcome. The maternal factors showed risks similar to the univariable (crude)

analysis. The addition of Big4 case mix adjustment (model 3) showed the intended place of

birth to be a significant co-variable, yet the contrast of planned home birth (OR 1.05 95%CI

0.91-1.21) vs. hospital birth (reference=1) turned out to be non-significant. The effect of

maternal risk factors was affected to a limited degree by the introduction of Big4 case mix.

We repeated the analysis for the PGA population (table 8.3). The results of the crude analysis

were close to the NPA analysis. However, the effect of ethnic background was considerably

stronger in the PGA population. In all analyses the intended place of birth showed a

consistent significant impact on intrapartum and early neonatal mortality, yet the contrast

between home and hospital birth never reached statistical difference. After Big4 case mix

adjustment home birth showed a non-significant increased risk (OR 1.11 95%CI 0.93-1.34).

Figure 8.3 describes the crude mortality risk (left Y-axis) and the Big4 adjusted home birth

mortality index (right Y-axis), where each dot represents the mortality risk results after the

group listed on the X-axis has been excluded from the population.

The crude mortality (dotted lines) initially shows a difference in favour of home delivery

(home: 0.18% vs. hospital: 0.22%), which converges towards a much lower average level if

premature births are excluded. Further exclusions lower the crude mortality rate, leaving the

small difference almost unaffected. The mortality index (black line) shows a distinct change

from an initial level of about 100% towards about 120% after exclusion of the pregnancy

138

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Ta

ble

8.1

Ch

arac

teris

tics

and

outc

omes

of

wom

en in

prim

ary

care

at

the

onse

t of

labo

ur (n

atur

al p

rosp

ectiv

e ap

proa

ch a

nd p

erfe

ct g

uide

line

appr

oach

)*

Varia

ble

Plan

ned

hom

e bi

rth

Plan

ned

hosp

ital b

irth

Plan

ned

plac

e un

know

nIn

trap

artu

m a

nd e

arly

ne

onat

al d

eath

NPA

PG

A

NPA

PG

A

NPA

PGA

N

PA

PGA

Parit

y†40

2,91

2 (5

9.3)

363,

568

(60.

4)21

9,10

5 (3

2.2)

190,

098

(31.

6)57

,935

(8.5

)48

,665

(8.1

)67

9,95

260

2,33

1Pr

imip

arou

s17

1,98

6 (4

2.69

)14

8,08

2 (4

0.73

)10

4,24

9 (4

7.58

)88

,110

(46.

35)

26,2

54 (4

5.32

)21

,047

(43.

25)

614

(0.2

0)28

3 (0

.11)

Mul

tipar

ous

230,

926

(57.

31)

215,

486

(59.

27)

114,

856

(52.

42)

101,

988

(53.

65)

31,6

81 (5

4.68

)27

,618

(56.

75)

485

(0.1

3)26

8 (0

.08)

Mat

erna

l age

(y)†

Youn

ger t

han

194,

036

(1.0

0)3,

502

(0.9

6)6,

713

(3.0

6)5,

770

(3.0

4)1,

190

(2.0

5) 9

10 (1

.87)

42 (0

.35)

13 (0

.13)

20-2

534

,661

(8.6

0)30

,787

(8.4

7)32

,617

(14.

89)

28,6

69 (1

5.08

)6,

823

(11.

78)

5,61

1 (1

1.53

)13

3 (0

.18)

65 (0

.10)

25-3

429

6,12

8 (7

3.50

)26

7,40

8 (7

3.55

)14

2,59

7 (6

5.08

)12

4,07

1 (6

5.27

)39

,526

(68.

22)

33,5

83 (6

9.01

)69

3 (0

.14)

348

(0.0

8)O

lder

than

35

68,0

87 (1

6.90

)61

,871

(17.

02)

37,1

78 (1

6.97

)31

,588

(16.

62)

10,3

96 (1

7.94

)8,

559

(17.

59)

231

(0.2

0)12

5 (0

.12)

Ethn

ic b

ackg

roun

d†

Wes

tern

364,

796

(90.

54)

329,

677

(90.

68)

143,

677

(65.

57)

124,

144

(65.

31)

45,2

05 (7

8.03

)38

,508

(68.

80)

880

(0.1

6)45

2 (0

.09)

Non

Wes

tern

38,1

16 (9

.46)

33,8

91 (9

.32)

75,4

28 (3

4.43

)65

,954

(34.

69)

12,7

30 (2

1.97

)17

,461

(31.

20)

219

(0.1

7)99

(0.0

8)

Nei

ghbo

urho

od†

Priv

ilege

d ne

igh-

bour

hood

388,

089

(96.

32)

350,

346

(96.

36)

196,

659

(89.

76)

170,

366

(89.

62)

53,8

23 (9

2.90

)45

,425

(93.

34)

1,03

1 (0

.16)

518

(0.0

9)

Und

erpr

ivile

ged

neig

hbou

rhoo

d14

,823

(3.6

8)13

,222

(3.6

4)22

,446

(10.

24)

19,7

32 (1

0.38

)4,

112

(7.1

0)3,

240

(6.6

6)68

(0.1

6)33

(0.0

9)

139

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Ges

tatio

nal a

ge (w

k)†

Less

than

34

2,40

9 (0

.60)

1,70

2 (0

.78)

583

(1.0

1)37

0 (7

.88)

35-3

66,

510

(1.6

2)4,

064

(1.8

5)1,

206

(2.0

8)65

(0.5

5)37

15,2

03 (3

.77)

13,6

22 (3

.75)

9,60

3 (4

.38)

8,46

8 (4

.45)

2,49

7 (4

.31)

2,18

7 (4

.49)

56 (0

.21)

51 (0

.21)

38-4

136

8,92

6 (9

1.56

)34

9,94

6 (9

6.25

)19

3,81

6 (8

8.46

)18

1,63

0 (9

5.55

)49

,585

(85.

59)

46,4

78 (9

5.51

)54

8 (0

.09)

500

(0.0

9)M

ore

than

41

9,86

4 (2

.45)

9,92

0 (4

.53)

4,06

4 (7

.01)

60 (0

.25)

Big4

†

Smal

l for

gest

atio

nal a

ge18

,786

(4.6

6)17

,089

(4.7

0)13

,114

(5.9

9)11

,604

(6.1

0)3,

081

(5.3

2)2,

665

(5.4

8)71

(0.2

0)59

(0.1

9)

Prem

atur

e8,

090

(2.0

1)5,

117

(2.3

4)1,

547

(2.6

7)92

(0.6

2)0.

00Lo

w A

pgar

scor

e1,

692

(0.4

2)1,

483

(0.4

1)1,

180

(0.5

4)95

9 (0

.50)

289

(0.5

0)22

8 (0

.47)

97 (3

.07)

86 (3

.22)

Cong

enita

l ab

nom

ality

4,87

4 (1

.21)

4,36

6 (1

.20)

2,94

1 (1

.34)

2,53

1 (1

.33)

778

(1.3

4)65

5 (1

.35)

74 (0

.86)

60 (0

.79)

Com

bina

tion

Big4

1,64

8 (0

.41)

693

(0.1

9)1,

279

(0.5

8)45

3 (0

.24)

391

(0.6

7)92

(0.1

9)45

8 (1

3.80

)85

(6.8

7)To

tal B

ig4

35,0

90 (8

.71)

23,6

31 (6

.50)

23,6

31 (1

0.79

)15

,547

(8.1

8)6,

086

(10.

50)

3,64

0 (7

.48)

792

(1.2

2)29

0 (0

.68)

NPA

, nat

ural

pro

spec

tive

appr

oach

; PG

A, p

erfe

ct g

uide

line

appr

oach

.D

ata

are

n (%

).* T

otal

s m

ay n

ot a

dd u

p to

100

bec

ause

of r

ound

ing

erro

r.† P

<.00

1.

140

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Ta

ble

8.2

In

trap

artu

m a

nd n

eona

tal d

eath

0-7

day

s in

wom

en w

ho a

re in

prim

ary

care

at t

he o

nset

of l

abou

r (na

tura

l pro

spec

tive

appr

oach

)

Mod

el 1

Mod

el 2

Mod

el 3

Tota

lM

orta

lity

pCr

ude

RR

95%

CI

pAd

just

ed

OR

95%

CI

pAd

just

ed

OR

95%

CI

p

Inte

nded

pla

ce o

f birt

h<.

05<.

05<.

05H

ome

402,

912

594

(0.1

5)0.

800.

71–0

.91

nie

1.05

0.91

–1.2

1H

ospi

tal [

ref]

219,

105

403

(0.1

8)1

1U

nkno

wn

57,9

3510

2 (0

.18)

0.96

0.77

–1.1

9ni

e0.

770.

61–0

.97

Parit

y <

.001

<.0

01 <

.001

Prim

ipar

ous

302,

489

614

(0.2

0)1.

581.

40–1

.78

1.67

1.47

–1.8

9ni

eM

ultip

arou

s [r

ef]

377,

463

485

(0.1

3)1

1

Mat

erna

l age

(y)

<.0

01 <

.001

<.0

01 <.

001

Youn

ger t

han

1911

,939

42 (0

.35)

2.43

1.78

–3.3

11.

801.

31–2

.48

1.67

1.17

–2.3

820

-25

74,1

0113

3 (0

.18)

1.24

1.03

–1.4

91.

030.

85–1

.24

0.92

0.75

–1.1

325

-34

[ref]

478,

017

693

(0.1

4)1

11

Old

er th

an 3

511

5,66

123

1 (0

.20)

1.38

1.19

–1.6

01.

561.

34–1

.81

1.44

1.23

–1.6

8

Ethn

ic b

ackg

roun

d <

.001

<.0

01<.

05<.

05W

este

rn [r

ef]

571,

185

880

(0.1

5)1

11

Non

-Wes

tern

108,

767

219

(0.2

0)1.

311.

13–1

.52

1.32

1.14

–1.5

41.

211.

02–1

.45

141

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Nei

ghbo

urho

od<.

05Pr

ivile

ged

neig

h-bo

urho

od [r

ef]

638,

571

1,03

1 (0

.16)

1

Und

erpr

ivile

ged

neig

hbou

rhoo

d41

,381

68 (0

.16)

1.02

0.80

–1.3

0ni

eni

e

Ges

tatio

nal a

ge (w

k) <

.001

<.0

01 <.

001

Less

than

34

4,69

437

0 (7

.88)

88.0

877

.44–

100.

1815

.74

12.5

6–19

.72

35-3

611

,780

65 (0

.55)

6.17

4.77

–7.9

72.

111.

56–2

.86

3727

,303

56 (0

.21)

2.29

1.74

–3.0

22.

261.

71–2

.99

38-4

1 [r

ef]

612,

327

548

(0.0

9)1

1M

ore

than

41

23,8

4860

(0.2

5)2.

812.

15–3

.67

2.10

1.60

–2.7

7

Big4

<.0

01 <

.001

<.00

1SG

A34

,910

71 (0

.20)

1.27

1.00

–1.6

23.

642.

82–4

.69

Prem

atur

ity14

,754

92 (0

.62)

12.4

99.

90–1

5.76

Low

apg

ar3,

161

97 (3

.07)

21.3

517

.28–

26.3

853

.41

42.5

5–67

.04

Cong

enita

l ab

nom

ality

8,59

374

(0.8

6)5.

684.

48–7

.20

14.5

311

.30–

18.6

7

Com

bina

tion

Big4

3,31

845

8 (1

3.80

)16

8.85

148.

94–1

91.4

168

.17

55.2

7–84

.08

No

Big4

615,

145

307

(0.0

5)1

RR, r

elat

ive

risk;

CI,

confi

den

ce in

terv

al; O

R, o

dds

ratio

; nie

, not

in e

quat

ion.

Dat

a ar

e n

or n

(%) u

nles

s ot

herw

ise

spec

ifi ed

.M

odel

1: c

rude

RR.

Mod

el 2

: adj

uste

d fo

r mat

erna

l fac

tors

incl

udin

g in

tend

ed p

lace

of b

irth,

par

ity, a

ge, e

thni

c ba

ckgr

ound

, and

nei

ghbo

rhoo

d. M

odel

3: a

djus

ted

for

mat

erna

l fac

tors

and

chi

ld fa

ctor

s; m

odel

2 +

ges

tatio

nal a

ge a

nd p

rese

nce

of B

ig4.

142

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Ta

ble

8.3

In

trap

artu

m a

nd n

eona

tal d

eath

0-7

day

s in

wom

en w

ho a

re in

prim

ary

care

at t

he o

nset

of l

abou

r (pe

rfec

t gui

delin

e ap

proa

ch)

Mod

el 1

Mod

el 2

Mod

el 3

Tota

l (n)

Mor

talit

yCr

ude

RR

95%

CI

pAd

just

ed

OR

95%

CI

pAd

just

ed

OR

95%

CI

p

Inte

nded

pla

ce o

f birt

h<.

05<.

05<.

05H

ome

363,

568

344

(0.0

9)0.

990.

83–1

.18

1.02

0.85

–1.2

31.

110.

93–1

.34

Hos

pita

l [re

f]19

0,09

818

2 (0

.10)

11

1U

nkno

wn

48,6

6525

(0.0

5)0.

540.

35–0

.81

0.54

0.36

–0.8

30.

570.

37–0

.86

Parit

y <.

001

<.00

1Pr

imip

arou

s25

7,23

928

3 (0

.11)

1.42

1.20

–1.6

71.

521.

28–1

.82

nie

Mul

tipar

ous

[ref

]34

5,09

226

8 (0

.08)

11

Mat

erna

l age

(y)

<.00

1 <.

001

<.05

Youn

ger t

han

1910

,182

13 (0

.13)

1.56

0.90

–2.7

11.

310.

75–2

.30

1.29

0.73

–2.2

620

-25

65,0

6765

(0.1

0)1.

220.

94–1

.59

1.11

0.84

–1.4

51.

080.

83–1

.41

25-3

4 [r

ef]

424,

915

348

(0.0

8)1

11

Old

er th

an 3

510

2,01

812

5 (0

.12)

1.50

1.22

–1.8

41.

661.

34–2

.04

1.50

1.22

–1.8

5

Ethn

ic b

ackg

roun

dW

este

rn [r

ef]

507,

063

452

(0.0

9)1

Non

-Wes

tern

94,7

1799

(0.1

0)1.

170.

94–1

.46

nie

nie

143

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

Nei

ghbo

urho

odPr

ivile

ged

neig

h-bo

urho

od [r

ef]

566,

137

518

(0.0

9)1

Und

erpr

ivile

ged

neig

hbou

rhoo

d36

,194

33 (0

.09)

1.00

0.70

–1.4

2ni

eni

e

Ges

tatio

nal a

ge (w

k) <.

001

<.00

137

24,2

7751

(0.2

1)2.

431.

82–3

.24

2.51

1.87

–3.3

738

-41

[ref]

578,

054

500

(0.0

9)1

1

Big3

<.00

1 <.

001

SGA

31,3

5859

(0.1

9)4.

033.

04–5

.35

4.28

3.22

–5.6

8Lo

w a

pgar

2,67

086

(3.2

2)69

.05

54.2

8–87

.84

71.5

255

.87–

91.5

6Co

ngen

ital

abno

mal

ity7,

552

60 (0

.79)

17.0

312

.88–

22.5

316

.76

12.6

4–22

.23

Com

bina

tion

Big3

1,23

885

(6.8

7)14

7.19

115.

97–1

86.8

015

9.49

123.

86–2

05.3

8N

o Bi

g355

9,51

326

1 (0

.05)

11

RR, r

elat

ive

risk;

CI,

confi

den

ce in

terv

al; O

R, o

dds

ratio

; nie

, not

in e

quat

ion.

Mod

el 1

: cru

de R

R. M

odel

2: a

djus

ted

for m

ater

nal f

acto

rs in

clud

ing

inte

nded

pla

ce o

f birt

h, p

arity

, age

, eth

nic

back

grou

nd, a

nd n

eigh

borh

ood.

Mod

el 3

: adj

uste

d fo

r m

ater

nal f

acto

rs a

nd c

hild

fact

ors;

mod

el 2

+ g

esta

tiona

l age

and

pre

senc

e of

Big

4.

144

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

duration <36 weeks. Combined with the similar crude mortality rates of home and hospital

delivery from then onwards, this suggests setting safety for the risk groups still included i.e.

all groups right to the exclusion label ‘pregnancy duration <36 weeks’. For example after

exclusion of pregnancy duration > 41 weeks (PGA group), the adjusted mortality index is

120%, which is slightly larger than the non significant regression result of 111% (table 8.3).

DISCUSSIONPlanned home birth within the Dutch maternity care system has a lower crude mortality

rate compared to a community midwife led planned hospital birth. However, after case

mix adjustment, the relation is reversed, showing a non-significant increased perinatal

mortality rate of home birth. Excess setting dependent mortality may arise at home if

risk conditions are present or emerge at birth, yet remnant confounding by indication

effect (Big4 conditions are more prevalent in hospital) and low mortality prevalence limits

statistical proof. Authors favouring a comparison of settings among ‘suitable’ home births

only (PGA), usually exclude risk conditions with a potential setting effect. This mechanism

may explain the apparently contradictory results from previous studies.1,5,7,10-15,17,18

Figure 8.3 ‘Big4’ adjusted mortality index of home birth (hospital based birth under midwife supervision=100%).

PGANPA

145

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

A strength of this study was the size of the study population, which reflects the complete

Dutch experience from 2000-2007. The amount of missing explanatory data is negligible,

mortality data have been shown to be complete. No annual trends are observed in the

relations shown, except for a minimal gradual decrease in total perinatal mortality.3

Our case mix adjustment proved to be essential. The assumption of comparability across

home vs. hospital populations appeared not to be justifiable judging from the unequal

prevalence of Big4 conditions. These primarily have their origin in early negative fetal

conditions and disadvantaged genetic background of the parents. Only in the case of low

Apgar, one may argue that the midwifery management during labour might influence it’s

occurrence, while a management role in SGA, spontaneous prematurity, and congenital

anomalies at that stage is unlikely. We decided to include low Apgar cases assuming the

role of management to be small compared to the disadvantage of the home setting once

a child with persistent low Apgar is born. Thus, our point of departure starts from the risk

challenge represented by Big4 at the onset of labour, and investigates whether setting

matters in terms of prognosis. The mechanisms underlying the apparent favourable

selection for home birth are still to be elucidated. Self selection by the pregnant women

can coincide with implicit or explicit selection by the midwife who may tend to ‘refer’ to

hospital if she feels uncomfortable with the risk level at home. The difference in the ratio

home:hospital community midwifery led deliveries among the four largest Dutch cities

suggests the presence of substantial professional and setting effects. In Amsterdam and

Utrecht the ratio is 2:1, and in Rotterdam and the Hague it is 1:2.

Several study limitations merit discussion. While an improvement compared to previous

studies, our case mix control is still incomplete because Big4 is unrelated to 15% of deaths.

In the PGA population this proportion is even 48%. Thus we cannot rule out remnant

confounding by indication as little is known on the factors underlying choice of setting.

RCT would be the superior design to address our research question. However when home

birth was part of a trial, participation hampered24 and introduced selective participation

which limited generalis ability. Moreover if following one’s choice impacts outcome,

estimates of setting effects are also biased.24-26 Despite their shortcomings, in particular

when considering the difficulty to overcome the confounding by indication phenomenon,

observational studies as ours are therefore invaluable. A comparison with a 100%

gynaecologist hospital-based system is not included. The data from an otherwise very

similar country as Flanders27 suggest that more favourable results may be expected in low

risk women in general from a hospital-based system. In Flanders perinatal mortality is about

33% less than in the Netherlands, while the caesarean section rates show little difference.

146

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

This study primarily focuses upon the disadvantages and neglects the claimed benefits

when comparing planned home versus planned hospital births. However studies accessing

mother’s opinion show that preventing these disadvantages easily outweighs the claimed

benefits.28

Our results appear compatible with most other reports even though previous studies show

conflicting results. Planned home births attended by registered professional attendants are

not associated with an increased risk of adverse perinatal outcomes in cohort studies in

North America7,12, the United Kingdom14, Europe 5,11,15,17, Australia29 and New Zealand30. In

contrast, other cohort studies have shown a higher risk of perinatal mortality in planned

home births compared to planned hospital births.10,13,16,18,30 All studies are limited by

voluntary submission of data7,8,11-14,17,31,32, non representative sampling5,13, lack of appropriate

comparison groups7,12,15,29, or insufficient statistical power5,17,29,32. A critical factor, as our

study shows, is the in retrospect exclusion of unplanned and unsuitable home births from

analysis.18

Our results partly agree with those of Kennare at al.30 who found higher standardised

perinatal mortality ratios among planned home deliveries after limited adjustment (birth

weight, gestational age). Our results also partly agree with the meta-analysis by Wax et al.9:

differences in the prevalence of SGA, premature births and congenital anomalies seem

equally present in planned home vs. hospital births. They reported a twofold higher neonatal

mortality rate but no increase in perinatal mortality. These results are in agreement with

figure 8.3 where the fetal death subgroup does not benefit from setting safety. It should be

noted that the study of Wax et al. received methodological criticisms33-36 most notably the

inclusion of the study of Pang and the exclusion of the study of De Jonge. Our conclusions

apparently contradict those of De Jonge et al. who concluded equal intrapartum and

early neonatal outcome of planned home birth vs. hospital birth in apparently the same

population.15 However, the point of departure is not the same. Of our two comparisons of

home delivery vs. hospital delivery, one parallels the approach of De Jonge. Our principal

approach (NPA) compares neonatal mortality in the actual populations delivering at home

vs. hospital, while the approach of De Jonge compares neonatal mortality in a hypothetical

group resembling our PGA population. Our adjustment procedure however goes further

than the maternal factor adjustment of De Jonge.15

From our study we conclude that planned home birth, under routine conditions, is not

associated with a higher intrapartum and early neonatal mortality rate. However in

subgroups additional risk cannot be excluded.

147

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

REFERENCES1. Home birth--proceed with caution. Lancet. 2010;

376:303.

2. Royal College of Obstetricians and Gynaecologists/Royal College of Midwives. Home births. Joint statement No.2; 2007 (Online: http://www.rcog.org.uk/files/rcog-corp/uploaded-files/JointStatmentHomeBirths2007.pdf).

3. Ravelli AC, Tromp M, van Huis M, et al. Decreasing perinatal mortality in The Netherlands, 2000-2006: a record linkage study. J Epidemiol Community Health. 2009;63:761-5.

4. CVZ. List of Obstetric Indications (in Dutch: Ver-loskundig Vademecum) 2003. Diemen (online: http://www.knov.nl/docs/uploads/Verloskundig_Vademecum_2003.pdf).

5. Ackermann-Liebrich U, Voegeli T, Gunter-Witt K, et al. Home versus hospital deliveries: follow up study of matched pairs for procedures and outcome. Zurich Study Team. BMJ. 1996;313:1313-8.

6. Hutton EK, Reitsma AH, Kaufman K. Outcomes associated with planned home and planned hospital births in low-risk women attended by midwives in Ontario, Canada, 2003-2006: a retrospective cohort study. Birth. 2009;36:180-9.

7. Janssen PA, Saxell L, Page LA, Klein MC, Liston RM, Lee SK. Outcomes of planned home birth with registered midwife versus planned hospital birth with midwife or physician. CMAJ. 2009;181:377-83.

8. Johnson KC, Daviss BA. Outcomes of planned home births with certified professional midwives: large prospective study in North America. BMJ. 2005;330:1416.

9. Wax JR, Lucas FL, Lamont M, Pinette MG, Cartin A, Blackstone J. Maternal and newborn outcomes in planned home birth vs planned hospital births: a metaanalysis. Am J Obstet Gynaecol. 2010;203:243 e1-8.

10. Wax JR, Pinette MG, Cartin A, Blackstone J. Maternal and newborn morbidity by birth facil-ity among selected United States 2006 low-risk births. Am J Obstet Gynaecol. 2010;202:152 e1-5.

11. Wiegers TA, Keirse MJ, van der Zee J, Berghs GA. Outcome of planned home and planned hospital births in low risk pregnancies: prospective study in midwifery practices in The Netherlands. BMJ. 1996;313:1309-13.

12. Anderson RE, Murphy PA. Outcomes of 11,788 planned home births attended by certified nurse-midwives. A retrospective descriptive study. J Nurse Midwifery. 1995;40:483-92.

13. Bastian H, Keirse MJ, Lancaster PA. Perinatal death associated with planned home birth in Australia: population based study. BMJ. 1998;317:384-8.

14. Chamberlain G, Wraight A, Crowley P. Birth at home. Pract Midwife. 1999;2:35-9.

15. De Jonge A, van der Goes BY, Ravelli AC, et al. Perinatal mortality and morbidity in a nationwide cohort of 529,688 low-risk planned home and hospital births. BJOG. 2009;116:1177-84.

16. Evers AC, Brouwers HA, Hukkelhoven CW, et al. Perinatal mortality and severe morbidity in low and high risk term pregnancies in the Netherlands: prospective cohort study. BMJ. 2010;341:c5639.

17. Lindgren HE, Radestad IJ, Christensson K, Hild-ingsson IM. Outcome of planned home births compared to hospital births in Sweden between 1992 and 2004. A population-based register study. Acta Obstet Gynaecol Scand. 2008;87:751-9.

18. Pang JW, Heffelfinger JD, Huang GJ, Benedetti TJ, Weiss NS. Outcomes of planned home births in Washington State: 1989-1996. Obstet Gynaecol. 2002;100:253-9.

19. Hildingsson IM, Lindgren HE, Haglund B, Radestad IJ. Characteristics of women giving birth at home in Sweden: a national register study. Am J Obstet Gynaecol. 2006;195:1366-72.

20. Hogberg U. Homebirths in a modern setting - a cautionary tale. Acta Obstet Gynaecol Scand. 2008;87:797-9.

21. Neuhaus W, Piroth C, Kiencke P, Gohring UJ, Mall-man P. A psychosocial analysis of women plan-ning birth outside hospital. J Obstet Gynaecol. 2002;22:143-9.

148

PLAN

NED

HO

ME CO

MPA

RED W

ITH PLA

NN

ED H

OSPITA

L BIRTHS

8

22. Perinatal care in The Netherlands 2006 (in Dutch: Perinatale zorg in Nederland 2006). Utrecht: The Netherlands Perinatal Registry; 2008.

23. Bonsel GJ, Birnie E, Denktas S, Poeran J, Steegers EAP. Dutch report: Lijnen in de Perinatale Sterfte, Signalementstudie Zwangerschap en Geboorte 2010. Rotterdam: Erasmus MC; 2010. (online: http://www.nvk.nl/Nieuws/Dossiers/DossierPerinatalezorg.aspx).

24. Hendrix M, Van Horck M, Moreta D, et al. Why women do not accept randomisation for place of birth: feasibility of a RCT in The Netherlands. BJOG. 2009;116:537-42; discussion 42-4.

25. Tyson H. Outcomes of 1001 midwife-attended home births in Toronto, 1983-1988. Birth. 1991;18: 14-9.

26. Dowswell T, Thornton JG, Hewison J, et al. Should there be a trial of home versus hospital delivery in the United Kingdom? BMJ. 1996;312:753-7.

27. Cammu H, Martens G, Landuyt J, De Coen K, Defoort P. Perinatale activiteiten in Vlaanderen 2009. Brussel: Studiecentrum voor Perinatale Epidemiologie; 2010.

28. Bijlenga D, Birnie E, Bonsel GJ. Feasibility, Reliability, and Validity of Three Health-State Valuation Methods Using Multiple-Outcome Vignettes on Moderate-Risk Pregnancy at Term. Value Health. 2009;12:821-7.

29. Woodcock HC, Read AW, Bower C, Stanley FJ, Moore DJ. A matched cohort study of planned home and hospital births in Western Australia 1981-1987. Midwifery. 1994;10:125-35.

30. Kennare RM, Keirse MJ, Tucker GR, Chan AC. Planned home and hospital births in South Australia, 1991-2006: differences in outcomes. Med J Aust. 2010;192:76-80.

31. Gulbransen G, Hilton J, McKay L, Cox A. Home birth in New Zealand 1973-93: incidence and mortality. N Z Med J. 1997;110:87-9.

32. Davies J, Hey E, Reid W, Young G. Prospective regional study of planned home births. Home Birth Study Steering Group. BMJ. 1996;313:1302-6.

33. Gyte GM, Dodwell MJ, Macfarlane AJ. Home birth metaanalysis: does it meet AJOG’s reporting re-quirements? Am J Obstet Gynaecol. 2011;204:e15; author reply e8-20, discussion.

34. Johnson KC, Daviss BA. International data demon-strate home birth safety. Am J Obstet Gynaecol. 2011;204:e16-7; author reply e8-20, discussion.

35. Kirby RS, Frost J. Maternal and newborn outcomes in planned home birth vs planned hospital births: a metaanalysis. Am J Obstet Gynaecol. 2011;204:e16; author reply e8-20, discussion.

36. Sandall J, Bewley S, Newburn M. “Home birth triples the neonatal death rate”: public commu-nication of bad science? Am J Obstet Gynaecol. 2011;204:e17-8; author reply e8-20, discussion.

9

Manuscript submitted for publication.

Does centralisation of acute obstetric care reduce perinatal

mortality? An empirical study of over 1 million births in

The Netherlands

J. Poeran, G.J.J.M. Borsboom, J.P. de Graaf, E. Birnie, E.A.P.Steegers, J.P. Mackenbach, G.J. Bonsel

150

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

ABSTRACTObjective Estimating outcome of hypothetical closure of 10 small hospitals (out of 99)

and centralisation of acute obstetric care in larger hospitals in order to assess whether this

would lower perinatal mortality in The Netherlands.

Design Hypothetical analysis using retrospective cohort data.

Setting The Netherlands

Population Selected were all (n=1,160,708) singleton hospital births from ≥22 weeks’

gestation as recorded in The Netherlands Perinatal Registry 2000-2008 with exclusion of (1)

unknown gestational age, (2) unknown or no travel time to hospital and (3) fetal mortality.

Main outcome measures Predicted perinatal (intrapartum and first-week) mortality for

several patient subgroups for two simulated centralisation scenarios: (1) closure of the

10 smallest hospitals, and (2) closure of the 10 smallest hospitals, but avoiding adjacent

closures. Women who delivered in hypothetically closed hospitals were assigned the next-

nearest hospital with travel time and hospital features changed accordingly. Predictions

followed from regression coefficients from a multilevel logistic regression model with a

forced casemix control (maternal and child factors) and a set of hospital organisational

features with perinatal mortality as outcome.

Results Scenario 1 resulted in doubled travel time, and 10% increased mortality (0.34% vs.

0.38%). In scenario 2 perinatal mortality showed little change (0.33% vs. 0.32%) with less

effect on travel time. Heterogeneity in hospital organisational features caused simultaneous

improvement and deterioration of predicted perinatal mortality depending mainly on the

features of the newly assigned next-nearest hospital. Consequences vary for subgroups:

pregnant women at increased risk suffer more from increased travelling, and gain more if

a (specialised) ‘perinatal centre’ or a better performing hospital is nearby.

Conclusion In The Netherlands, centralisation of acute obstetric care according to the

‘closure-of-the-smallest-rule’ yields suboptimal outcomes. In order to develop an optimal

strategy for centralisation one would need to consider all positive and negative effects:

heterogeneity in organisational features of closed and surviving hospitals, financial aspects,

differential effects for patient subgroups, increased travel time and adequate antenatal

risk selection.

151

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

INTRODUCTIONDutch perinatal mortality exceeds the European average.1,2 Structural inadequacies in the

provision of care were discovered as major contributors.3 In response, a national Steering

Group on Pregnancy and Birth -on behalf of the Dutch Ministry of Health-, issued several

recommendations.4 One recommendation addressed the observed inadequate availability

of 7*24h acute obstetric care stating availability within 15 minutes of ‘qualified professionals’

(midwives, gynaecologists, paediatricians, anaesthesiologists, and operating theatre staff;

‘qualified’ in terms of seniority).4,5 However, small hospitals reported to be unable to satisfy

these demands with existing on-call coverage schemes. Moreover, larger hospitals were

unwilling to do so as the current financial system explicitly excludes any reimburse ment of

so-called ‘availability (standby) costs’. Consequently, centralisation of acute obstetric care

services was considered, implying a reduction of acute services in about half of all hospitals

with a parallel redistri bution of qualified professionals from small local to nearby larger

hospitals. While the advantages are obvious (e.g., increase in continuity of care, and a likely

volume related performance improvement), the disadvantages of centralisation should

be acknowledged too (e.g., increased travel time and organisational disutilities regarding

non-obstetric services).4-9 Centralisation may also critically affect paediatric, emergency

room, and acute anaesthesia services in the reduced hospitals.8

The key question is whether direct effects of centralisation on patient outcomes are to be

expected. An overall positive outcome would more easily justify the above disutilities. In

this study we estimated the direct effects on intrapartum and first-week mortality when

10 small hospitals (out of 99 providing obstetric care) would hypothetically be closed

according to two plausible centralisation scenarios. The effects of closing were estimated by

a redistribution of patient flow (depending on their zip code of residency) to the next-nearest

hospital, taking into account (1) detailed information of maternal and child characteristics,

(2) acute referral status, (3) travel time in normal and (in retrospect) high risk cases, and

hospital related factors such as (4) day and time of birth, (5) the hospital’s organisational

7*24h characteristics, and (6) any additional non-specific hospital effects not accounted

for by the previously mentioned factors. The net consequences were then calculated for all

individual women redistributed from the hypothetically closed hospitals to the remaining

hospitals, to allow for a trade-off of positive and negative effects.

152

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

METHODS

General approach

This study was issued by the so-called society of SAZ-hospitals, which represents the

interests of about 40 small general hospitals in The Netherlands. These hospitals are

located throughout The Netherlands, either within the ‘Randstad’ conurbation (the densely

populated western part of the Netherlands) or outside. Centralisation of care primarily will

affect SAZ hospitals.

Our general approach was as follows. In step 1, we estimated the impact of maternal,

child, and hospital’s organisational characteristics in terms of intrapartum and first-week

(perinatal) mortality, using multilevel logistic regression analysis. Next, in step 2, we

identified all pregnant women who would need to change hospital as the consequence of

a defined centralisation scenario. We distinguished two centralisation scenarios: closure of

the 10 smallest hospitals (scenario 1) and closure of the 10 smallest hospitals, while avoiding

simultaneous closure of two adjacent hospitals (scenario 2). The women involved were

assumed to deliver in the next-nearest hospital in terms of estimated travel time. In step

3, we estimated the individual intrapartum and first-week mortality risk after closure. We

did so by applying to each individual woman the regression coefficients (obtained in step

1) to the characteristics of her situation after centralisation. Finally, in step 4, we compared

the current observed total mortality among women affected by closure, with the expected

total mortality by adding the estimated probabilities.

Data collection

Data on maternal factors, child factors and outcomes were obtained from The Netherlands

Perinatal Registry. This registry contains population-based information of 97% of all

pregnancies in the Netherlands.10 On behalf of the respective professional societies,

source data are routinely collected by 94% of midwives, 99% of gynaecologists and 68%

of paediatricians including 100% of Neonatal Intensive Care Unit (NICU) paediatricians

(see website for detailed description: www.perinatreg.nl).10 The professional board of The

Netherlands Perinatal Registry granted permission to use the anonymous registry data for

this study. The registry contains 1,620,126 records for the period 2000-2008. We excluded

multiple pregnancies (n=35,326), births with unknown gestational age (n=17,278), births with

an unknown or erroneous zip code (n=23,736), births in hospitals which had not participated

in the registry >2 years (n=6,535), homebirths (n=368,672), and stillbirths (n=7,871).

153

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

We excluded homebirths since centralisation of care, travel distance, and hospital’s

organisational features are of little or no importance in homebirths. Stillbirths were excluded

as delivery decisions are subject to different considerations.11

Analysed were 1,160,708 births which additionally were combined with data from two

sources: travel time data and individual hospital’s organisational characteristics. Travel

times to the closed and next-nearest hospitals were calculated making use of the hospital’s

zip code and pregnant women’s zip code (these 4-digit zip codes cover on average

4,000 inhabitants). Travel times (not: distances) between zip codes were derived from a

commercially available geographic information system (www.geodan.nl). We distinguished

planned from unplanned births as we assume travel time to hospital only to have an impact

on women with unplanned births who were transferred from home to hospital during

parturition. Thus, hospital travel time was set to zero for women with planned births, i.e.,

induction of labour or a planned (primary) caesarean section.

Data on hospital’s organisational characteristics including data pertinent to 7*24 hour

services were collected separately from this study, with the support of the Dutch Society of

Obstetricians and Gynaecologists using a standard questionnaire for all 99 Dutch maternity

units (response: 100%).3

Table 9.1 displays these collected 24 variables with inevitable overlap of data. We used

principal component analysis (PCA) to summarise the original 24 variables into two principal

components, factor 1 and factor 2, with a computed numerical score for each hospital.

Factor 1 can be interpreted as ‘hospital’s scale size’ as it primarily combines original variables

pointing to hospital size and the number of obstetric caregivers: the lower the factor value

the larger the hospital and the higher the number of obstetric caregivers. Factor 2 can be

interpreted as ‘24-hour equality of service level’. This factor combines original variables

pointing to around-the-clock availability of the various qualified professionals: the higher

the factor value the more the 24-hour equality of service level (in terms of seniority). These

two factors jointly explained almost 70% of variance of the original 24 variables, i.e., the

two constructed variables contain 70% of the available information.

PCA is a ‘data reduction’ technique commonly used in datasets with a high number of related

explanatory variables. It essentially summarises the net information content of overlapping

data into a small number of independent constructed variables. The background and the

technique itself are described in more depth elsewhere.12

154

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Ta

ble

9.1

O

verv

iew

of h

ospi

tals

’ org

anis

atio

nal d

ata

Org

anis

atio

nal f

eatu

reRe

mar

ks

Gen

eral

Teac

hing

hos

pita

lca

tego

rised

into

yes

/ no

Gyn

aeco

logy

dep

artm

ent

Dur

atio

n of

day

time

shift

sca

tego

rised

into

dur

atio

ns o

f 10-

12 h

ours

and

7-9

hou

rsD

urat

ion

of e

veni

ng /

nigh

t shi

fts

and

the

high

est l

evel

of t

he

prof

essi

onal

who

is a

tten

ding

thes

e sh

ifts

cate

goris

ed in

to d

urat

ions

of ≤

14 h

ours

and

≥15

hou

rs w

ith p

rofe

ssio

nal l

evel

s ca

tego

rised

in

to tw

o gr

oups

: (1)

gyn

aeco

logi

st /

gyna

ecol

ogis

t in

trai

ning

and

(2) n

on-in

-tra

inin

g ph

ysic

ian

/ mid

wife

/ nu

rse

The

high

est l

evel

of t

he p

rofe

ssio

nal w

ho is

att

endi

ng th

e ev

enin

g an

d ni

ght s

hift

s du

ring

the

wee

kca

tego

rised

into

fi ve

gro

ups:

(1) g

ynae

colo

gist

, (2)

in-t

rain

ing

and

non-

in-t

rain

ing

resi

dent

gy

naec

olog

y, (3

) gen

eral

em

erge

ncy

phys

icia

n, (4

) mid

wife

and

(5) n

urse

The

high

est l

evel

of t

he p

rofe

ssio

nal w

ho is

att

endi

ng th

e da

ytim

e sh

ifts

durin

g w

eeke

nds

cate

goris

ed in

to fi

ve g

roup

s: (1

) gyn

aeco

logi

st, (

2) in

-tra

inin

g an

d no

n-in

-tra

inin

g re

side

nt

gyna

ecol

ogy,

(3) g

ener

al e

mer

genc

y ph

ysic

ian,

(4) m

idw

ife a

nd (5

) nur

seTh

e hi

ghes

t lev

el o

f the

pro

fess

iona

l who

is a

tten

ding

the

even

ing

and

nigh

t shi

fts

durin

g w

eeke

nds

cate

goris

ed in

to fi

ve g

roup

s: (1

) gyn

aeco

logi

st, (

2) in

-tra

inin

g an

d no

n-in

-tra

inin

g re

side

nt

gyna

ecol

ogy,

(3) g

ener

al e

mer

genc

y ph

ysic

ian,

(4) m

idw

ife a

nd (5

) nur

sePe

rmitt

ed to

sle

ep d

urin

g at

tend

ing

shift

sca

tego

rised

into

yes

/ no

Gyn

aeco

logi

st o

n ba

ckup

cal

l dur

ing

shift

sca

tego

rised

into

yes

/ no

Prof

essi

onal

doi

ng ro

unds

dur

ing

wee

kend

sca

tego

rised

into

four

gro

ups:

(1) g

ynae

colo

gist

, (2)

in-t

rain

ing

and

non-

in-t

rain

ing

resi

dent

gy

naec

olog

y, (3

) gen

eral

em

erge

ncy

phys

icia

n, (4

) mid

wife

and

(5) n

urse

In-h

ospi

tal p

rese

nce

of th

e em

erge

ncy

oper

atin

g th

eatr

e te

amca

tego

rised

into

thre

e gr

oups

: in-

hosp

ital p

rese

nce

durin

g (1

) 24

hour

s, (2

) pre

senc

e du

ring

dayt

ime

and

even

ing

and

on-c

all d

urin

g th

e ni

ght,

(3) p

rese

nce

durin

g da

ytim

e an

d on

-cal

l du

ring

the

even

ing

and

nigh

t In

-hos

pita

l pre

senc

e of

an

anae

sthe

siol

ogis

tca

tego

rised

into

thre

e gr

oups

: in-

hosp

ital p

rese

nce

durin

g (1

) 24

hour

s, (2

) pre

senc

e du

ring

dayt

ime

and

even

ing

and

on-c

all d

urin

g th

e ni

ght,

(3) p

rese

nce

durin

g da

ytim

e an

d on

-cal

l du

ring

the

even

ing

and

nigh

t Th

e to

tal n

umbe

r of o

bste

tric

car

egiv

ers

(spe

cial

ists

, ph

ysic

ians

, mid

wiv

es, n

urse

s)a

cont

inuo

us n

umbe

r

Ann

ual n

umbe

r of d

eliv

erie

sco

ntin

uous

as

wel

l as

grou

ped

into

thre

e ca

tego

ries:

(1) <

1000

, (2)

100

0-20

00, (

3) >

2000

155

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Com

bina

tion

of th

e an

nual

num

ber o

f del

iver

ies

with

the

gyna

ecol

ogis

t:oth

er o

bste

tric

car

egiv

er ra

tioca

tego

rised

into

six

gro

ups:

(1) <

1000

and

≤1:

4, (2

) <10

00 a

nd 1

:4-1

:6, (

3) <

1000

and

>1:

6, (4

) 10

00-2

000

and

≤1:4

, (5)

100

0-20

00 >

1:4,

(6) >

2000

and

all

ratio

s

Min

imum

dur

atio

n of

att

endi

ng s

hift

dur

ing

wee

kend

sa

cont

inuo

us n

umbe

r (ra

nge

12-7

2 ho

urs)

Aver

age

trav

el ti

me

in m

inut

es fr

om h

ome

to h

ospi

tal f

or a

n on

-cal

l gyn

aeco

logi

sta

cont

inuo

us n

umbe

r

Max

imum

trav

el ti

me

in m

inut

es fr

om h

ome

to h

ospi

tal f

or a

n on

-cal

l gyn

aeco

logi

sta

cont

inuo

us n

umbe

r

Num

ber o

f car

egiv

ers

pres

ent d

urin

g at

tend

ing

shift

s on

w

eekd

ay e

veni

ngs

and

nigh

tsa

cont

inuo

us n

umbe

r

Num

ber o

f car

egiv

ers

pres

ent d

urin

g at

tend

ing

shift

s on

w

eeke

nds

durin

g th

e da

ya

cont

inuo

us n

umbe

r

Num

ber o

f car

egiv

ers

pres

ent d

urin

g at

tend

ing

shift

s on

w

eeke

nds

durin

g th

e ev

enin

g an

d th

e ni

ght

a co

ntin

uous

num

ber

Paed

iatr

ics

depa

rtm

ent

Leve

l of c

are

for n

ewbo

rns

cate

goris

ed in

to fo

ur g

roup

s: (1

) Neo

nata

l Int

ensi

ve C

are

Uni

t (N

ICU

), (2

) pos

t-N

ICU

or

incu

bato

r fro

m 3

0 w

eeks

of g

esta

tiona

l age

, (3)

incu

bato

r fro

m 3

2 w

eeks

of g

esta

tiona

l age

, (4

) ini

tially

fi rs

t aid

for n

eona

te w

ith s

ubop

timal

sta

rt w

ith s

ubse

quen

t ref

erra

l to

a ho

spita

l w

ith a

hig

her l

evel

of c

are

The

high

est l

evel

of t

he p

rofe

ssio

nal a

t the

pae

diat

rics

depa

rtm

ent w

ho is

att

endi

ng th

e sh

ifts

outs

ide

offi c

e ho

urs

cate

goris

ed in

to s

ix g

roup

s: (1

) fel

low

neo

nato

logi

st, (

2) p

aedi

atric

ian,

(3) i

n-tr

aini

ng

paed

iatr

icia

n, (4

) non

-in-t

rain

ing

resi

dent

pae

diat

rics,

(5) g

ener

al e

mer

genc

y ph

ysic

ian,

(6) n

o pr

ofes

sion

al p

rese

nt b

ut o

n-ca

ll Th

e lo

wes

t lev

el o

f the

pro

fess

iona

l at t

he p

aedi

atric

s de

part

men

t who

is a

tten

ding

the

shift

s ou

tsid

e offi

ce

hour

sca

tego

rised

into

six

gro

ups:

(1) f

ello

w n

eona

tolo

gist

, (2)

pae

diat

ricia

n, (3

) in-

trai

ning

pa

edia

tric

ian,

(4) n

on-in

-tra

inin

g re

side

nt p

aedi

atric

s, (5

) gen

eral

em

erge

ncy

phys

icia

n, (6

) no

prof

essi

onal

pre

sent

but

on-

call

Prof

essi

onal

on-

call

at th

e pa

edia

tric

s de

part

men

t ca

tego

rised

into

four

gro

ups:

(1) n

o on

e on

-cal

l, (2

) pae

diat

ricia

n on

-cal

l, (3

) fel

low

ne

onat

olog

ist o

n-ca

ll, (4

) oth

er p

rofe

ssio

nal o

n-ca

ll

156

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Outcome and determinants of hospital performance

The main outcome was intrapartum or first-week mortality after live birth, for convenience

further referred to as perinatal mortality. Determinants with their optimal categorisation

were selected for their proven effect on perinatal mortality.3,7,13-17 We distinguished maternal

and child factors (casemix) and organisational factors. Maternal factors were parity and

age category combined (primiparous / multiparous, <25 / 25-29 / 30-34 / 35-39 / ≥40

years), and ethnicity (Western / non-Western). Child factors were gestational age (thirteen

categories3), congenital anomalies (yes / no), small for gestational age (SGA: birthweight

>10th / 2.3-10th / <2.3rd percentile) and fetal presentation (cephalic / breech / transverse

or other / unknown). Congenital anomalies are recorded postpartum and classified through

a standard coding system by organ system (8 categories, 71 subcategories).18

Organisational factors were travel time to the hospital in minutes (continuous), the two

principal components (factor 1 and factor 2), day and time of delivery (Saturday / Sunday /

weekday, each subdivided into three time slots 00:00-07:59 / 08:00-17:59 / 18:00-00:00), and

emergency referral during parturition (yes / no). We assumed the hospitals’ performance

according to principal component factor 1 and 2 did not to change after the hypothetical

closure of the 10 small hospitals. Emergency referral during parturition was defined by

using information from both the caregiver during pregnancy (midwife, obstetrician) and

the caregiver responsible at start of labour (midwife or obstetrician).

Analysis

For presentational purposes only, we grouped hospitals into small size SAZ hospitals (1)

within and (2) outside the Randstad conurbation, (3) academic hospitals and (4) other

general hospitals.

The different steps in the analysis are illustrated in figure 9.1. First, a comprehensive

multilevel multivariable logistic regression model (‘hospital performance model’) was

fitted using the primary dataset (step 1). All determinants mentioned above were entered

into the model. The principal component factors were included in the model as both their

actual value as well as their quadratic value as we presumed linear effects as well as effects

of extremes (e.g., the largest hospitals’ performance to be comparable to the smallest

hospitals in which quality in the former is ensured by the total staffing levels [‘quantity’]

and in the latter by the level of obstetric staff [mostly gynaecologists, not residents on

duty, ‘quality’])

157

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

We predefined an additional interaction term relating travel time to clinical condition of

the child, as previous studies have suggested a poorer condition to be associated with a

larger adverse impact of extended travel time.19 Clinical condition was represented by the

single or combined presence of so-called ‘Big4’20 morbidities: congenital anomalies, preterm

birth (<37th week of gestation), SGA (birthweight <10th percentile for gestational age) or

low Apgar score (<7, 5 minutes after birth). As the presence of any of the ‘Big4’ morbidities

precedes perinatal mortality in 85% of cases, we assume this to be an appropriate risk

indicator.20 Because of the multilevel structure of the data (individuals are grouped per

hospital) we used multilevel logistic regression analysis which adjusts for the possible

clustering of individuals within hospitals, and unspecified effects at the hospital level. In

step 2 women who delivered in the hypothetically closed hospitals were assigned the

next-nearest hospital. Finally, the estimated coefficients obtained from step 1 were used to

predict perinatal mortality for each scenario (step 3) taking into account (1) the features of

the new hospital, (2) the associated travel distance and (3) the specific interaction term of

‘Big4’*travel time for those referred. The overall observed and predicted mortality before

vs. after the closure of the hospitals were then compared. Predicted mortality was the sum

of all individually recalculated mortality probabilities for women affected by centralisation.

Figure 9.1 Overview of computational model.

PRIMARY DATASET DUPLICATE DATASETS

DATASET 1 / SCENARIO 1‘Closure of 10 smallest hospitals’

-women in closed hospitals assigned the next nearest hospital with corresponding travel time and hospital features

FITTING OF HOSPITAL PERFORMANCE MODEL

PREDICTING PERINATAL MORTALITYwith coefficients from hospital performance model

DATASET 2 / SCENARIO 2‘Closure of 10 smallest hospitals, avoiding adjacent closures’

-women in closed hospitals assigned the next nearest hospital with corresponding travel time and hospital features

158

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Presentation was for all women together, and for several subgroups based on planned

or unplanned birth, referral during parturition, deliveries within ‘office hours’, parity, and

presence of morbidities. The statistical software package SAS version 9.2 (SAS Institute Inc.,

Cary, NC) was used, with the GLIMMIX procedure to run random intercept multilevel models.

RESULTSTable 9.2 shows the study population characteristics by type of hospital. Most women are

multiparous and 30-34 years (20.2-22.8%), and of Western origin (79.0-90.4%). Preterm birth

(<37 weeks’ gestation) was observed in 0.6-3.8%, congenital anomalies in 1.8-5.0%, SGA

in 9.4-11.3%, and a non-cephalic presentation in 7.2-9.2%. Of all women 2.1-12.9% had to

travel more than 20 minutes to the hospital where they gave birth. SAZ hospitals all fell

in the ≥ P25 value range of factor 1 (‘hospital scale size, number of obstetric caregivers’,

a lower value represents larger hospitals) while academic hospitals all fell in the <P25

value range. This confirms the SAZ hospitals to be relatively small with smaller numbers of

obstetric caregivers compared to the larger academic hospitals. Factor 2 (‘24-hour equality

of service level’ a higher factor value represents more 24-hour equality of service level in

terms of seniority) was distributed more evenly across type of hospital. The majority of

women delivered in hospitals in the ≥ P25 value range illustrating most women to deliver

in hospitals with the highest level of 24-hour equality of service. Most deliveries were on a

weekday between 08:00 and 17:59 (37.9-43.9%); 22.2-29.9% of women were referred during

parturition. ‘Big4’ risk prevalence did not differ much between SAZ hospitals (17.3-19.2%)

and other non-academic hospitals (19.0%). For academic hospitals prevalence is 27.3%.

Hospital performance

Effect estimates from the multilevel multivariable logistic regression model (‘hospital

performance model’) are shown in table 9.3. All casemix variables are significantly associated

with perinatal mortality. The highest risks are seen for multiparous women of 35-39 years

(OR 1.51 CI 1.30-1.75), non-Western women (OR 1.15 CI 1.04-1.28), gestational ages <37 (OR

range 2.97->1000), congenital anomalies (OR 10.11 CI 9.14-11.19), birthweight <p2.3 (OR

6.46 CI 5.67-7.35), and transverse or other fetal presentation (OR 2.06 CI 1.60-2.65).

For organisational factors only principal component factor 1, 2 and quadratic factor 2 are

not significantly associated with perinatal mortality. Travel time (in minutes) to hospital has

an OR of 0.98, implying a 2% decreased risk of perinatal mortality for every extra minute of

159

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Ta

ble

9.2

Ch

arac

teris

tics

of th

e st

udy

popu

latio

n pe

r hos

pita

l gro

up

SAZ

hosp

itals

with

inRa

ndst

adco

nurb

atio

n

SAZ

hosp

itals

outs

ide

Rand

stad

conu

rbat

ion

Acad

emic

hos

pita

lsO

ther

hos

pita

ls

N%

N%

N%

N%

Ca

sem

ix v

ari

ab

les

Parit

y &

age

cat

egor

yPr

imip

arou

s &

<25

yea

rs9,

600

9.2%

22,7

379.

6%13

,654

9.0%

67,3

2410

.1%

Prim

ipar

ous

& 2

5-29

yea

rs18

,477

17.7

%47

,239

19.9

%23

,546

15.5

%11

8,19

717

,7%

Prim

ipar

ous

& 3

0-34

yea

rs19

,113

18.3

%37

,042

15.6

%25

,826

17.0

%11

6,58

917

.5%

Prim

ipar

ous

& ≥

35 y

ears

7,46

27.

1%10

,945

4.6%

11,7

557.

7%42

,888

6.4%

Mul

tipar

ous

& <

25 y

ears

2,58

22.

5%6,

805

2.9%

4,81

43.

2%21

,252

3.2%

Mul

tipar

ous

& 2

5-29

yea

rs10

,270

9.8%

29,1

4812

.3%

15,8

2910

.4%

71,6

3610

.7%

Mul

tipar

ous

& 3

0-34

yea

rs21

,067

20.2

%54

,102

22.8

%31

,184

20.5

%13

5,85

020

.4%

Mul

tipar

ous

& 3

5-39

yea

rs13

,551

13.0

%25

,978

10.9

%21

,153

13.9

%79

,350

11.9

%M

ultip

arou

s &

≥40

yea

rs2,

369

2.3%

3,63

11.

5%4,

232

2.8%

13,5

112.

0%

Ethn

icity

Wes

tern

83,8

6580

.3%

214,

887

90.4

%12

0,03

679

.0%

529,

076

79.4

%N

on-W

este

rn20

,626

19.7

%22

,740

9.6%

31,9

5721

.0%

137,

521

20.6

%

Ges

tatio

nal a

ge22

-27.

6 w

eeks

127

0.1%

348

0.1%

3,19

02.

1%1,

317

0.2%

28-3

1.6

wee

ks11

90.

1%39

30.

2%6,

605

4.3%

1,42

10.

2%32

wee

ks24

80.

2%71

30.

3%1,

296

0.9%

2,05

10.

3%33

wee

ks53

90.

5%1,

361

0.6%

1,28

00.

8%3,

653

0.5%

34 w

eeks

944

0.9%

2,36

31.

0%1,

804

1.2%

6,23

80.

9%

Tabl

e 9.

1 co

ntin

ues o

n ne

xt p

age.

160

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Ta

ble

9.2

Co

ntin

ued

SAZ

hosp

itals

with

inRa

ndst

adco

nurb

atio

n

SAZ

hosp

itals

outs

ide

Rand

stad

conu

rbat

ion

Acad

emic

hos

pita

lsO

ther

hos

pita

ls

N%

N%

N%

N%

35 w

eeks

1,55

61.

5%3,

967

1.7%

2,51

71.

7%10

,363

1.6%

36 w

eeks

2,94

92.

8%7,

255

3.1%

4,64

13.

1%19

,611

2.9%

37 w

eeks

6,77

16.

5%15

,563

6.5%

9,48

76.

2%39

,044

5.9%

38 w

eeks

16,3

4215

.6%

38,8

1416

.3%

21,7

8414

.3%

97,3

5714

.6%

39 w

eeks

23,1

5722

.2%

52,2

6222

.0%

30,9

1720

.3%

147,

476

22.1

%40

wee

ks27

,052

25.9

%58

,970

24.8

%35

,281

23.2

%17

0,67

225

.6%

41 w

eeks

18,5

8217

.8%

41,3

7017

.4%

24,1

7115

.9%

119,

727

18.0

%≥

42 w

eeks

6,10

55.

8%14

,248

6.0%

9,02

05.

9%47

,667

7.2%

Cong

enita

l ano

mal

ies

No

102,

648

98.2

%22

9,99

796

.8%

144,

320

95.0

%65

1,15

097

.7%

Yes

1,84

31.

8%7,

630

3.2%

7,67

35.

0%15

,447

2.3%

Smal

l for

ges

tatio

nal a

ge (S

GA

)N

o SG

A94

,585

90.5

%21

4,31

290

.2%

134,

787

88.7

%59

6,67

189

.5%

Birt

hwei

ght P

2.3-

P10

7,66

37.

3%17

,706

7.5%

12,9

368.

5%53

,289

8.0%

Birt

hwei

ght <

P2.

32,

243

2.1%

5,60

92.

4%4,

270

2.8%

16,6

372.

5%

Feta

l pre

sent

atio

nCe

phal

ic97

,009

92.8

%21

8,01

091

.7%

137,

933

90.7

%61

8,14

892

.7%

Bree

ch5,

873

5.6%

15,3

606.

5%10

,822

7.1%

40,0

996.

0%Tr

ansv

erse

or o

ther

720

0.7%

1,83

90.

8%1,

510

1.0%

5,03

30.

8%U

nkno

wn

889

0.9%

2,41

81.

0%1,

728

1.1%

3,31

70.

5%

161

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Org

an

isa

tio

na

l fa

cto

rs

Trav

el ti

me

to h

ospi

tal

≤ 20

min

utes

102,

320

97.9

%21

7,04

891

.3%

132,

426

87.1

%62

7,18

694

.1%

> 20

min

utes

2,17

12.

1%20

,579

8.7%

19,5

6712

.9%

39,4

115.

9%

Hos

pita

l org

anis

atio

nal f

eatu

res

Prin

cipa

l com

pone

nt (f

acto

r 1)

≥ P2

510

4,49

110

0.0%

237,

627

100.

0%0

0.0%

539,

523

80.9

%<

P25

00.

0%0

0.0%

151,

993

100.

0%12

7,07

419

.1%

Prin

cipa

l com

pone

nt (f

acto

r 2)

≥ P2

587

,162

83.4

%23

7,62

710

0.0%

151,

993

100.

0%40

0,36

860

.1%

< P2

517

,329

16.6

%0

0.0%

00.

0%26

6,22

939

.9%

Day

and

tim

e of

del

iver

ySa

turd

ay 0

0:00

-07:

593,

432

3.3%

7,70

43.

2%5,

985

3.9%

24,1

953.

6%Sa

turd

ay 0

8:00

-17:

595,

485

5.2%

13,1

775.

5%8,

248

5.4%

37,1

695.

6%Sa

turd

ay 1

8:00

-23:

592,

846

2.7%

6,45

82.

7%4,

397

2.9%

19,3

582.

9%Su

nday

00:

00-0

7:59

3,38

63.

2%7,

814

3.3%

5,97

73.

9%24

,464

3.7%

Sund

ay 0

8:00

-17:

595,

736

5.5%

12,8

465.

4%8,

259

5.4%

37,1

185.

6%Su

nday

18:

00-2

3:59

2,76

82.

6%6,

209

2.6%

4,44

72.

9%19

,329

2.9%

Wee

kday

s 00

:00-

07:5

917

,600

16.8

%39

,607

16.7

%30

,296

19.9

%12

5,36

318

.8%

Wee

kday

s 08

:00-

17:5

945

,837

43.9

%10

3,15

243

.4%

57,5

4437

.9%

266,

601

40.0

%W

eekd

ays

18:0

0-23

:59

17,4

0116

.7%

40,6

6017

.1%

26,8

4017

.7%

113,

000

17.0

%

Refe

rral

dur

ing

part

uriti

onN

o76

,046

72.8

%17

1,32

572

.1%

118,

186

77.8

%46

7,60

270

.1%

Yes

28,4

4527

.2%

66,3

0227

.9%

33,8

0722

.2%

198,

995

29.9

%

Big4

Cat

egor

yN

o Bi

g486

,454

82.7

%19

2,10

380

.8%

110,

463

72.7

%53

9,70

881

.0%

Onl

y on

e Bi

g416

,724

16.0

%41

,741

17.6

%33

,636

22.1

%11

6,16

017

.4%

Mor

e th

an o

ne B

ig4

1,31

31.

3%3,

783

1.6%

7,89

45.

2%10

,729

1.6%

162

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Table 9.3 Hospital performance model

aOR CI P

Casemix variables

Parity & age category <.0001Primiparous & <25 years 1.09 0.93 1.28Primiparous & 25-29 years [ref ] 1.00Primiparous & 30-34 years 1.09 0.95 1.26Primiparous & ≥35 years 1.32 1.11 1.58Multiparous & <25 years 0.88 0.68 1.14Multiparous & 25-29 years 1.32 1.13 1.55Multiparous & 30-34 years 1.45 1.27 1.65Multiparous & 35-39 years 1.51 1.30 1.75Multiparous & ≥25 years 1.51 1.16 1.97

Ethnicity 0,009Western [ref ] 1.00Non-Western 1.15 1.04 1.28

Gestational age <.000122-27.6 weeks >1,000 >1,000 >1,00028-31.6 weeks 68.41 55.92 83.6732 weeks 14.64 11,18 19.1733 weeks 11.94 9.52 14.9734 weeks 6.23 4.95 7.8435 weeks 4.88 3.96 6.0036 weeks 2.97 2.44 3.6137 weeks 2.16 1.82 2.5838 weeks 1.18 1.00 1.3939 weeks 1.05 0.90 1.2240 weeks [ref ] 1.0041 weeks 1.38 1.18 1.60≥ 42 weeks 1.37 1.11 1.69

Congenital anomalies <.0001No [ref ] 1.00Yes 10.11 9.14 11.19

Small for gestational age (SGA) <.0001No SGA [ref ] 1.00Birthweight P2.3-P10 2.51 2.24 2.83Birthweight < P2.3 6.46 5.67 7.35

Fetal presentation <.0001Cephalic [ref ] 1.00Breech 1.65 1.46 1.85Transverse or other 2.06 1.60 2.65Unknown 1.48 1.02 2.13

Table 9.3 continues on next page.

163

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Table 9.3 Continued

aOR CI P

Organisational factors

Travel time (to hospital, in minutes) 0.98 0.97 0.99 <.0001

Hospital organisational featuresPrincipal component (factor 1) 0.99 0.87 1.13 0,893Principal component (factor 2) 1.08 0.92 1.27 0,356Principal component (factor 1) ^2 0.85 0.72 1.00 0,048Principal component (factor 2) ^2 1.04 0.92 1.18 0,510

Day and time of delivery <0.001Saturday 00:00-07:59 1.50 1.22 1.85Saturday 08:00-17:59 1.17 0.97 1.40Saturday 18:00-23:59 1.41 1.12 1.78Sunday 00:00-07:59 1.31 1.06 1.63Sunday 08:00-17:59 1.46 1.24 1.73Sunday 18:00-23:59 1.33 1.05 1.68Weekdays 00:00-07:59 1.35 1.21 1.51Weekdays 08:00-17:59 [ref ] 1.00Weekdays 18:00-23:59 1.40 1.25 1.56

Referral during parturition <0.001No [ref ] 1.00Yes 1.28 1.17 1.41

Interaction <.0001No Big4*travel time [ref ] 1.00Only one Big4*travel time 0.99 0.98 0.99More than one Big4*travel time 1.02 1.01 1.02

aOR: adjusted odds ratio, CI: confi dence interval.

travel time. Deliveries on Saturday nights have a 50% increased risk for perinatal mortality

(OR 1.50; CI 1.22-1.85), referral during parturition has a 28% increased risk (OR 1.28; CI 1.17-

1.41). The interaction term Big4*travel time is ‘protective’ in case of one ‘Big4’ morbidity

(OR 0.99; CI 0.98-0.99), but complex ‘Big4’ morbidity adds to the travel risk (RR 1.02; CI 1.01-

1.02). The random coefficient for hospital varied (data not shown) which is suggestive of

differences in hospital performance in terms of perinatal mortality.

Eff ects of centralisation scenarios

Table 9.4 and table 9.5 list the predicted results of the centralisation scenarios 1 and 2. The

left side of the table (‘before closure’) depicts the observed situation while the right side

(‘after closure’) depicts the predicted effects for the defined subgroups.

164

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Scenario 1

Following scenario 1 (table 9.4), a total number of 61,578 (5.3%) women will have to deliver

in other hospitals as their hospital of (actual) delivery will be hypothetically closed. Perinatal

mortality is expected to increase with 21 (10%) cases: from 210 cases (0.34%) observed

before closure to 231 cases (0.38%) predicted after closure. In absolute numbers, expected

perinatal mortality after closure is higher than the observed mortality before closure for

almost all subgroups except for unplanned hospital births (from 124 to 116 mortality cases)

Table 9.4 Predicted results after hypothetical closure of 10 hospitals according to scenario 1

Group Before closure After closure

N † N † % N † N † %

All pregnant women 61,578 210 0.34% 61,578 231 0.38%

Planned hospital birth (primary caesarean section/induced labour)

26,431 86 0.33% 26,431 115 0.44%

Unplanned hospital birth 35,147 124 0.35% 35,147 116 0.33%

Referral during parturition from home

6,775 19 0.28% 6,775 22 0.32%

Referral during parturition from hospital

7,599 20 0.26% 7,599 25 0.33%

Deliveries within 'offi ce hours' 26,618 75 0.28% 26,618 79 0.30%Deliveries outside 'offi ce hours' 34,960 135 0.39% 34,960 152 0.43%

Primiparous women 30,664 102 0.33% 30,664 113 0.37%Multiparous women 30,914 108 0.35% 30,914 118 0.38%

No Big4 morbidities 49,247 30 0.06% 49,217 58 0.12%Only one Big4 morbidity 11,175 91 0.81% 11,175 89 0.80%More than one Big4 morbidities 1,156 89 7.70% 1,156 84 7.27%

Minutes 95% CI Minutes 95% CI

Average travel time in unplanned hospital births

12.7 3.2-22.2 23.7 12.5-34.9

Travel time >20 minutes in hospital births

6,077 9.9% 20,201 32.8%

Hospital size (deliveries per year) N N

<750 per year 61,578 23,884750-1,250 per year 0 28,2731,250-1,750 per year 0 9,400>1,750 per year 0 21

†: perinatal mortality.

165

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

and the presence of one (from 91 to 89 mortality cases) or more (from 89 to 84 mortality

cases) ‘Big4’ morbidities. The largest difference in observed versus predicted perinatal

mortality is for the group of planned hospital births (from 86 to 115 mortality cases). In

addition, average travel time to hospital will increase from 12.7 to 23.7 minutes with almost

one third (32.8%) of women having to travel more than 20 minutes after closure. Less than

half of women (23,884 of 61,578) will deliver in an alternative hospital that is of the same

size group (<750 annual deliveries) as before closure.

Table 9.5 Predicted results after hypothetical closure of 10 hospitals according to scenario 2

Group Before closure After closure

N † N † % N † N † %

All pregnant women 81,852 268 0.33% 81,852 259 0.32%

Planned hospital birth (primary caesarean section/induced labour)

39,172 124 0.32% 39,172 122 0.31%

Unplanned hospital birth 42,680 144 0.34% 42,680 137 0.32%

Referral during parturition from home

11,113 28 0.25% 11,113 29 0.26%

Referral during parturition from hospital

11,037 22 0.20% 11,037 28 0.25%

Deliveries within 'offi ce hours' 36,285 95 0.26% 36,285 95 0.26%Deliveries outside 'offi ce hours' 45,567 173 0.38% 45,567 164 0.36%

Primiparous women 42,085 131 0.31% 42,085 119 0.28%Multiparous women 39,767 137 0.34% 39,767 140 0.35%

No Big4 morbidities 67,233 52 0.08% 67,233 86 0.13%Only one Big4 morbidity 13,488 99 0.73% 13,488 94 0.70%More than one Big4 morbidities 1,131 117 10.34% 1,131 79 6.98%

Minutes 95% CI Minutes 95% CI

Average travel time in unplanned hospital births

10.8 4.2-17.3 15.8 10.3-21.4

Travel time >20 minutes in hospital births

3,106 3.8% 9,167 11.2%

Hospital size (deliveries per year) N N

<750 per year 81,852 9,089750-1,250 per year 0 69,2041,250-1,750 per year 0 3,533>1,750 per year 0 26

†: perinatal mortality.

166

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Scenario 2

According to scenario 2 (table 9.5), a total number of 81,852 (7.1%) women will have to

deliver in other hospitals with the relocation resulting in a negligible effect on perinatal

mortality (from 268 to 259 cases). For most subgroups, expected perinatal mortality after

closure is lower than the observed mortality before closure, except for referral during

parturition from home (from 28 to 29 cases) or hospital (from 22 to 28 cases), deliveries

within ‘office hours’ (95 cases before and after closure), multiparous women (from 137 to

140 cases), and no ‘Big4’ morbidities (from 52 to 86 cases). In addition, average travel time

to hospital will increase from 10.8 to 15.8 minutes. Note that a minority of women (9,089

of 81,852, 11.1%) will deliver in an alternative hospital of the same size group (<750 annual

deliveries) as before closure.

DISCUSSION

Principal fi ndings

In this national study with retrospective cohort data we predicted outcome effects for two

centralisation scenarios for acute obstetrical care by using a comprehensive empirical model.

Combined with increased travel time to hospital for both scenarios, scenario 1 resulted

in a 10% overall increased perinatal mortality, while scenario 2 resulted in a negligible

effect on perinatal mortality. In scenario 2 >85% of women in hypothetically ‘closed’

hospitals were ‘transferred’ to higher volume hospitals. Rather than just increased travel

time, the effects of centralisation appear to be determined by heterogeneity of hospitals’

organisational characteristics which in turn increase or decrease predicted perinatal

mortality. These changes appeared to depend more on the features of the newly assigned

next-nearest hospital rather than that of the initial hypothetically closed hospital. Indirectly

and by their quality factor scores, the data demonstrate that small hospitals perform

reasonably well. The picture is subtle: risk groups, e.g., women with one or more ‘Big4’

morbidities benefit from both centralisation scenarios as their predicted perinatal mortality

is lower than observed perinatal mortality before closure, Conversely, centralisation

appears to be detrimental for the small subset of women who are referred during

parturition.

167

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

Other studies

Centralisation of acute obstetric care services has been considered in other countries as

well, mostly relying on theoretical reasoning alone.21-23 Furthermore, previous studies at

best used an ecological study design, were descriptive rather than comparative, and were

heterogeneous in design and results.9,19,22-24 Most studies conclude a strong pro- or against-

centralisation policy comparing risk-adjusted outcome in small (low level care) hospitals

versus outcome in large (high level care) hospitals, without modelling hospital performance

itself.9,19,22-25 In Norway, Moster et al. performed a population-based study using data on

1.7 million births from The Norwegian Medical Birth Registry.24 Neonatal mortality was

compared between several geographical areas characterised by the volume of the majority

of the maternity units. Overall neonatal mortality was 1.2-2.2 times higher in areas with the

majority of women giving birth in small scale maternity units (500 or less births annually)

compared to areas with the majority of births in large scale units (>3,000 births annually).

As this study lacked risk and casemix adjustment, its conclusion in favour of centralisation

may be biased in the presence of area related risk differences as shown in other studies.24

Some studies focus on either high risk or low risk patient groups.9,19,23,25 Bartels et al.25 and

Phibbs et al.19 studied the effect of NICU level on neonatal mortality for high risk infants,

both showing (up to 94%) higher (adjusted) neonatal mortality in smaller and lower level

NICUs favouring centralisation. Finally, two studies which focused on low risk patients show

conflicting results.9,23

Strengths and weaknesses

An important strength of our study is the usage of a validated national database as

virtually all pregnancies over 9 years were included (2000-2008). Additionally, our analysis

included detailed information on organisational characteristics of all Dutch hospitals,

lacking in other studies.9,19,22-25 Other strengths include the detailed casemix adjustment

and the multilevel approach which adjusts for the possible clustering of individuals within

hospitals. The hospital performance model combined with our scenario approach enables

detailed effect estimations of different hypothetical centralisation scenarios going beyond

mortality impact estimations only.9,19,22-25 Studies based on reasoning can account less

easily for the mix of perinatal mortality effects, and the subdivision of effects into effects in

different patient groups. Based on these strengths, our study goes beyond similar previous

studies’ methodology providing empirical evidence on differential effects of centralisation

schemes. We believe our framework to be beneficial for other countries also considering

168

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

centralisation of acute obstetric care. In the United Kingdom, for example, the Royal

College of Obstetricians and Gynaecologists recently cautioned the widespread belief

that centralisation will always result in better patient outcome by stating ‘localised where

possible, centralised where necessary’.21

This study has limitations. Maternal lifestyle factors, in particular smoking, were unavailable

and not included in the casemix adjustment26 By including SGA and gestational age in

the model the effects of this limitation (in the context of this study) are reduced as they, if

combined, act as a proxy for smoking. Smoking has little effect beyond its effect on birth

weight and gestational age at delivery. Furthermore, we based our estimations on the

assumption that women in the hypothetically closed hospitals would deliver in the next-

nearest hospital. While this strategy is generally valid, individual preferences may decide

otherwise.27 Also, we assumed no change in performance and organisational features of

remaining hospitals after centralisation. As conflicting study results illustrate9,19,22-24, it is

unclear in what direction these features will change after centralisation. Additional research

is needed.

Finally, we only included perinatal mortality, travel time and volume of destination hospitals

as effect measures of the centralisation scenarios to illustrate the differential effects of two

different centralisation scenarios. Future research with additional outcomes such as NICU-

admission or intervention rates may further specify the effects of centralisation.

Arguments pro and con centralisation

Common arguments in favour of centralisation are: volume-related better care provision,

access to rapid intervention during delivery, quick resuscitation of the newborn, and rapid

identification and management of newborn infants with unexpected morbidities (e.g.,

congenital anomalies) in large and higher level hospitals.24 An important reason opposing

centralisation is the increased travel time to hospital, with inherent increased risk for

adverse perinatal outcome, especially for high risk women or out-of-hospital delivery.22

Ravelli et al. showed a 17-52% increased risk for perinatal mortality with travel time to

hospital of 20 minutes or more.7 Another reason against centralisation is that in high risk

oriented hospitals, low risk births are expected to be less ‘natural’ with an increased risk for

(obstetric) interventions and corresponding increased costs.28 Moreover, financial aspects

of centralisation policy need to be considered, e.g., resource distribution costs or education

and salary costs for consultants (obstetrics, paediatrics, anaesthesiology, etc.).29 The inter-

dependency of acute obstetric care services with, e.g., paediatric / neonatal services and

169

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

anaesthesiology services should also be considered. These services may be at risk of collapse

as the number of physicians required to sustain on-call coverage may be no longer available

in small hospitals after acute obstetric care has been centralised.8 Finally, particularly for The

Netherlands where about 20% of births still occur at home under supervision of a midwife10,

increased travel time to hospital will compromise these (low risk) women in their choice

to opt for a birth at home.

Conclusion

In The Netherlands, centralisation of obstetric care according to the ‘closure-of-the-smallest-

rule’ yields suboptimal outcomes in terms of perinatal mortality and increased travel time

to hospital. In order to develop an optimal strategy for centralisation one would need to

consider all positive and negative effects: heterogeneity in organisational features of closed

and surviving hospitals, financial aspects, differential effects for patient subgroups, increased

travel time and adequate antenatal risk selection. Current heterogeneity in organisational

features, where the smallest hospitals on average perform remarkably well, contradicts a

simple size related closure strategy. Differential effects of hospital quality, travel time and

antenatal risk selection should all be taken into account.

REFERENCES1. Buitendijk SE, Nijhuis JG. High perinatal mortality

in the Netherlands compared to the rest of Europe. Ned Tijdschr Geneeskd. 2004;148:1855-1860.

2. Mohangoo AD, Buitendijk SE, Hukkelhoven CW, et al. Higher perinatal mortality in The Netherlands than in other European countries: the Peristat-II study. Ned Tijdschr Geneeskd. 2008;152:2718-2727.

3. de Graaf JP, Ravelli AC, Visser GH, et al. Increased adverse perinatal outcome of hospital delivery at night. BJOG. 2010;117:1098-1107.

4. Sheldon T. Prompt access to expert care is among advice to reduce perinatal deaths in Netherlands. BMJ. 2010;340:c277.

5. Sheldon T. Netherlands drops plans to deliver acute obstetric care in 15 minutes. BMJ. 2012;344: e2302.

6. 24/7 acute obstetric hospital care (in Dutch: 24/7 acute verloskunde in het ziekenhuis), Dutch report on the effects of centralisation of acute obstetric care. Amstelveen: KPMG Advisory Plexus; 2012.

7. Ravelli AC, Jager KJ, de Groot MH, et al. Travel time from home to hospital and adverse perinatal outcomes in women at term in the Netherlands. BJOG. 2011;118:457-465.

8. Klein M, Johnston S, Christilaw J, Carty E. Mothers, babies, and communities. Centralizing maternity care exposes mothers and babies to complications and endangers community sustainability. Can Fam Physician. 2002;48:1177-1179, 1183-1175.

9. Tracy SK, Sullivan E, Dahlen H, Black D, Wang YA, Tracy MB. Does size matter? A population-based study of birth in lower volume maternity hospitals for low risk women. BJOG. 2006;113:86-96.

170

CENTRA

LISATION

OF ACU

TE OBSTETRIC CA

RE IN TH

E NETH

ERLAN

DS

9

10. Perinatal care in The Netherlands 2008 (in Dutch: Perinatale zorg in Nederland 2008). Utrecht: The Netherlands Perinatal Registry; 2011.

11. Stillbirth Collaborative Research Network Writ-ing G. Causes of death among stillbirths. JAMA. 2011;306:2459-2468.

12. Burstyn I. Principal component analysis is a powerful instrument in occupational hygiene inquiries. Ann Occup Hyg. 2004;48:655-661.

13. Anthony S, Jacobusse GW, van der Pal-de Bruin KM, Buitendijk S, Zeitlin J, Factors E-PWGoR. Do differences in maternal age, parity and multiple births explain variations in fetal and neonatal mortality rates in Europe?--Results from the EURO-PERISTAT project. Paediatr Perinat Epidemiol. 2009;23:292-300.

14. De Graaf JP, Ravelli AC, Wildschut HI, et al. Perinatal outcomes in the four largest cities and in deprived neighbourhoods in The Netherlands. Ned Tijdschr Geneeskd. 2008;152:2734-2740.

15. Evers AC, Brouwers HA, Hukkelhoven CW, et al. Perinatal mortality and severe morbidity in low and high risk term pregnancies in the Netherlands: prospective cohort study. BMJ. 2010;341:c5639.

16. Ravelli AC, Tromp M, van Huis M, et al. Decreasing perinatal mortality in The Netherlands, 2000-2006: a record linkage study. J Epidemiol Community Health. 2009;63:761-765.

17. Timmermans S, Bonsel GJ, Steegers-Theunissen RP, et al. Individual accumulation of heterogeneous risks explains perinatal inequalities within deprived neighbourhoods. Eur J Epidemiol. 2011;26:165-180.

18. Rietberg CCT. Term breech delivery in The Netherlands; chapter 7. Thesis 2006, University of Utrecht. (online available: http://igitur-archive.library.uu.nl/dissertations/2006-1031-200813/c7.pdf ).

19. Phibbs CS, Bronstein JM, Buxton E, Phibbs RH. The effects of patient volume and level of care at the hospital of birth on neonatal mortality. JAMA. 1996;276:1054-1059.

20. Bonsel GJ, Birnie E, Denktas S, Poeran J, Steegers EAP. Dutch report: Lijnen in de Perinatale Sterfte, Signalementstudie Zwangerschap en Geboorte 2010. Rotterdam: Erasmus MC; 2010.

21. Tomorrow’s specialist: working party report, September 2012. London: The Royal College of Obstetricians and Gynaecologists; 2012. (Online http://www.rcog.org.uk/files/rcog-corp/Tomorrow’s%20Specialist_FullReport.pdf ).

22. Hemminki E, Heino A, Gissler M. Should births be centralised in higher level hospitals? Experiences from regionalised health care in Finland. BJOG. 2011;118:1186-1195.

23. Heller G, Richardson DK, Schnell R, Misselwitz B, Kunzel W, Schmidt S. Are we regionalized enough? Early-neonatal deaths in low-risk births by the size of delivery units in Hesse, Germany 1990-1999. Int J Epidemiol. 2002;31:1061-1068.

24. Moster D, Lie RT, Markestad T. Neonatal mortality rates in communities with small maternity units compared with those having larger maternity units. BJOG. 2001;108:904-909.

25. Bartels DB, Wypij D, Wenzlaff P, Dammann O, Poets CF. Hospital volume and neonatal mortality among very low birth weight infants. Pediatrics. Jun 2006;117:2206-2214.

26. Andres RL, Day MC. Perinatal complications associated with maternal tobacco use. Semin Neonatol. 2000;5:231-241.

27. Phibbs CS, Mark DH, Luft HS, et al. Choice of hospital for delivery: a comparison of high-risk and low-risk women. Health Serv Res. 1993;28:201-222.

28. Tracy SK, Tracy MB. Costing the cascade: estimat-ing the cost of increased obstetric intervention in childbirth using population data. BJOG. 2003; 110:717-724.

29. Lemmens K, Janus J, van Schooten G, Vlieger E. 24/7 acute obstetric care in hospital (in Dutch: 24/7 acute verloskunde in het ziekenhuis). Amstelveen: KPMG Plexus; 2012. (Online http://www.rijksoverheid.nl/documenten-en-publicaties/rapporten/2012/03/01/24-7-acute-verloskunde-in-het-ziekenhuis.html).

10

Manuscript submitted for publication.

The impact of extremes in outdoor temperature and sunshine

exposure on birthweight

J. Poeran, E. Birnie, E.A.P. Steegers, G.J. Bonsel

172

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

ABSTRACTBackground With the advent of the ‘fetal origins of adult disease’ hypothesis, environmental

determinants of birthweight regained interest. However, studies are heterogeneous in

design with conflicting results. We developed and applied a detailed spatial-time exposure

model for the most likely climatological factors affecting fetal growth: seasonality,

temperature and sunshine

Methods Daily climatological data (29 weather stations) were linked to all 1,460,401 term

singleton births (The Netherlands Perinatal Registry 2000-2008). An individual exposure

matrix for each pregnancy was computed for five exposure windows: (1) periconceptional,

(2) first / (3) second / (4) third trimester, and (5) day of delivery. Next, linear regression models

were specified with birthweight (in grams) as outcome and seasonality, minimum and

maximum ambient temperature, cumulative sunshine exposure, and maternal and child

factors as determinants. The final model enabled to quantify climatological contribution

to existing spatial variations in birthweight.

Results In The Netherlands, substantial regional differences exist in temperature extremes

and sunshine, with less temperature extremes and more sunshine in coastal areas. Our

exposure model explained existing irregular and modest seasonal birthweight effects

as a combined effect of season, temperature and sunshine with biologically plausible

exposure effects. A seasonal birthweight pattern emerged, modified by short and long

term temperature and sunshine effects. Summer is associated with an almost 20 gram

significantly decreased birthweight. Short term minimum and maximum temperature

exposures are significantly associated with higher and lower birthweight, respectively,

with effect sizes dependent on their timing during pregnancy. Also, higher cumulative

sunshine exposure during pregnancy increases birthweight. On the population level, we

demonstrated spatial differences in birthweight (range -11 to +25 grams) attributable to

the cumulative climatological effects, with lowest birthweights in inland areas.

Conclusion Birthweight is associated with the combined effect of season and exposure

to temperature extremes and cumulative exposure to sunshine, in particular during

critical reproductive exposure windows in pregnancy. The demonstrated negative

birthweight effects of maximum temperature exposure confirm results from animal studies.

Consequently, a climate footprint is visible in the regional birth weight differences.

173

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

INTRODUCTIONOver 2,000 years ago, Hippocrates already postulated effects of ‘warm and cold winds,

seasons and changes in weather’ on health.1 Indeed, a large body of evidence exists on

the effects of seasonality, ambient temperature and sunshine exposure on adult health.2-4

With the advent of the ‘fetal origins of adult disease’ hypothesis5, many studies have also

focused on exposure to climatological factors in the intrauterine period.6 In particular, the

determinants of birthweight are of renewed interest in the search to identify modifiable

factors which may prove useful for the development of strategies that prevent possible

disease in later life.6-10

Seasonality, ambient temperature and sunshine appear the most likely factors to influence

birthweight and have been previously studied.6-10 However, these studies show inconsistent

results, due to differences in methodology and study size, little or no adjustment for

known confounders, and considerable differences in exposure definitions.6-10 So far, no

universal exposure concept has emerged, distinguishing between peak exposure and

cumulative exposure effects, and taking varying exposure windows during pregnancy into

account. Most studies use mean temperature during pregnancy as exposure measure6-10,

while in particular extremes during heat and cold waves are of interest as they increase

cardiovascular mortality in adult deaths through blood flow changes.11 Regarding sunshine,

cumulative rather than peak exposure is of interest as it is thought that its effect runs mainly

through vitamin D.12

Here, we study the impact of peak temperature and cumulative sunshine on birthweight,

distinguishing between five exposure windows: periconceptionally, in the first, second

and third trimesters of pregnancy, and the day of delivery. We obtained geographically

detailed climatological data which permitted us to compute day-to-day individual exposure

patterns. The impact of climatological factors on birthweight was then studied using an

individual exposure model, taking into account the association between temperature and

sunshine exposure, and the competing impact of other known endogenous and exogenous

factors influencing birthweight.13 The resulting temporal-spatial relations were mapped,

to demonstrate the contribution of the selected climatological factors to existing regional

differences in birthweight.

174

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

METHODS

Patient data

In this retrospective cohort study we derived data on maternal factors, child factors and

birthweight from The Netherlands Perinatal Registry. This registry contains complete

population-based information of >97% of all pregnancies in the Netherlands.14 On behalf of

the respective professional societies, source data are routinely collected by 94% of midwives,

99% of gynaecologists and 68% of paediatricians including 100% of Neonatal Intensive

Care Unit paediatricians.14 (See website for detailed description: www.perinatreg.nl). The

board of The Netherlands Perinatal Registry granted permission to use the anonymous

registry data for this study. The registry contains data on 1,620,126 births for the period of

2000-2008. We excluded multiple pregnancies (n=35,326), births with gestational age < 37

weeks or unknown (n=116,139), births with unknown parity or neonate’s sex (n=353 and

n=433, respectively), and unknown or erroneous zip code (n=7,474). The final database

consisted of 1,460,401 births.

Climatological data

Daily maximum and minimum temperatures (in degrees Celsius) and sunshine (in hours)

were derived from 29 temperature stations throughout The Netherlands (figure 10.1). The

catchment area of each temperature station was based on 2-digit zip code with an average

population of 182,228 per zip code area. Data were obtained from the Royal Netherlands

Meteorological Institute (see website for detailed description: www.knmi.nl/index_en.html).

Exposure concept

From past evidence and biological considerations (critical reproductive windows) 6-10, we

derived five non-overlapping exposure windows (figure 10.2):

• from 3 days before to 3 days after conception, periconceptional (E1);

• from conception to 91 days after conception, first trimester (E2);

• from 92 days to 182 days after conception, second trimester (E3);

• from 183 days to 273 days after conception, third trimester (E4);

• day of birth (E5).

175

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

Climatological exposure was defined as follows: seasonality (captured by month of birth),

the minimum and maximum temperature, individually calculated for each exposure

window (E1-E5), and total sunshine exposure (in hours) individually calculated for windows

E2-E4.

The computation of exposure started with determining date of conception, which was

estimated from recorded information on gestational age at birth and date of delivery.

Figure 10.1 Map of The Netherlands with the 29 temperature stations and their catchment areas in shades of grey.

176

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

The climatological exposure information was subsequently projected on each pregnancy

individually for E1 to E5, providing an individual exposure profile for each woman. Climate

data were linked to individual records using the zip code. Date of delivery (a consecutive

number from 1 to 3,288 for every day in the study period) was used as variable to capture

the secular trend of increasing birthweight.8

Defi nition of outcome and maternal and child factors

The main outcome was birthweight in grams. Known maternal and child factors affecting

birthweight were also included. Maternal factors were parity (primiparous / multiparous),

ethnicity (Western / non-Western), maternal age (continuous), and socioeconomic status

(SES) score based on zip code of residence. Since the effect of maternal age on birthweight

appears to be inverse U-shaped (lower birthweight for the youngest and oldest mothers),

we used the quadratic value of maternal age as determinant. As a proxy for neighbourhood

Figure 10.2 Exposition model with different windows of exposure and the selection of exposure measures included the regression model. C = conception, C-3 = conception minus 3 days, C+91/182/273 = conception plus 91/182/273 days, E1-E5 = windows of exposure.

x xx x xx x xx x xx x

C+91 C+182 C+273

E2: from conception to 91 days after conceptionE3: from 92 days to 182 days after conceptionE4: from 183 days to 273 days after conception

C

E1E4

E5

C-3

E3E2

Total sunshine exposure in hours

MEASURE OF EXPOSURE

E5: day of birth

WINDOW OF EXPOSURE Minimum temperature

Maximum temperature

E1: 3 days before to 3 days after conception

177

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

SES we used zip code specific (publicly available) SES scores, which are made available by the

National Statistics Office (website in Dutch: www.scp.nl/Organisatie/Onderzoeksgroepen/

Wonen_Leefbaarheid_Veiligheid/Lopend_onderzoek_van_WLV/Statusscores). These SES

scores are zip code based and use individual data on e.g., income, taxes, hours of work, and

educational level. The scores are approximately normally distributed at the neighbourhood

level, where a negative score represents a high SES.

Child factors were gestational age (six categories), presence of congenital anomalies (yes

/ no), and neonate’s sex (male / female). In The Netherlands Perinatal Registry, congenital

anomalies are recorded postpartum and classified through a standard coding system by

organ system (8 categories, 71 subcategories).

Analysis

Linear regression models were specified in a predefined order. First, all maternal and child

factors and consecutive number of day of birth were included as determinants (model 0).

Next, season (month of birth) was added (model 1). In model 2 and 3, the temperature

and sunshine exposure measures were subsequently added. Models’ goodness of fit was

compared by the adjusted R-squared (R2) statistic.

Subsequently, we aimed to study regional differences in mean birthweight attributable to

our climatological exposure measures. We did so by studying the birthweight difference

between a dataset in which we allowed for the actual regional climatological differences

(original dataset) and a dataset in which we (hypothetically) ‘eliminated’ this regional

difference by substituting the minimum and maximum temperature, and sunshine

exposition by the national average, all other things equal (duplicate dataset). We used

coefficients obtained from model 3 to predict mean birthweight in this duplicate dataset.

The resulting regional differences (in grams) between observed mean weight (including

the effect of climatological factors) and predicted mean weight (excluding the effect of

climatological factors) are attributable to climatological factors and were illustrated on a

map of The Netherlands per 2-digit zip code. Because of the cancellation of the effect of

climatological factors in the duplicate dataset, the difference in mean weight between the

original and duplicate dataset represents the regional differences in birthweight attributable

to the selected climatological factors.

The linear regression models were fitted using SPSS version 20 (IBM Corporation, Somers,

NY, USA). SAS version 9.2 (SAS Institute Inc., Cary, NC) was used, with the GLM procedure

to run model 3 and use its coefficients to predict birthweight in a duplicate dataset. Maps

178

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

were constructed using ESRI ArcGIS version 9.3 (Environmental Systems Research Institute,

Inc., USA).

Table 10.1 Study group characteristics: subgroup sizes, subgroup mean birthweight differences with overall mean birthweight, and subgroup SGA prevalence (%)

N Mean subgroup BW-mean overall BWa

SD % SGA

Maternal factors

ParityPrimiparous 660,797 -88 485 9.7%Multiparous 799604 73 500 9.8%

Maternal age<20 years 23,594 -197 469 14.1%20-35 years 1,226,233 -2 497 9.6%>35 years 210,574 32 516 10.3%

EthicityWestern 1,225,392 18 500 9.0%Non-Western 235,009 -92 490 13.5%

Socioeconomic statusb

Low (<p20) 365,558 -56 504 12.2%Medium (p20-p80) 842,290 15 499 9.2%High (>p80) 252,553 33 490 8.2%

Child factors

Gestational age37 weeks 87,505 -473 488 9.8%38 weeks 221,349 -237 468 9.8%39 weeks 361,621 -67 450 9.8%40 weeks 428,778 85 453 9.7%41 weeks 281,509 210 466 9.8%≥42 weeks 79,639 282 476 9.7%

Congenital anomaliesNo 1,430,003 2 498 9.6%Yes 30,398 -97 581 16.1%

SexMale 746,603 67 505 9.7%Female 713,798 -70 485 9.8%

Year of birth2000-2002 501,459 -14 501 10.3%2003-2005 488,010 5 501 9.7%2006-2008 470,932 9 497 9.1%

a Mean overall BW = 3,522 gram, with SD = 500 gram.b Status scores were cut off at the 20th (low SES) and 80th percentile (high SES).

179

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

RESULTSTable 10.1 illustrates study population characteristics, differences between subgroup mean

birthweight and overall mean birthweight, and small for gestational age prevalence per

subgroup (SGA, birthweight <10th percentile). Subgroups with a significantly lower mean

birthweight are: (1) women who are primiparous (-88 grams), <20 years (-197 grams), non-

Western origin/background (-92 grams), with a low socioeconomic status (-56 grams); and

children who are born between 37 and 39 weeks of gestational age (-473 to -67 grams),

who have congenital anomalies (-97 grams), and are female (-70 grams). SGA prevalence

was highest in teenage pregnancies (maternal age <20 years, 14.1%), non-Western women

(13.5%), low socioeconomic status women (12.2%), and children born with a congenital

anomaly (16.1%).

Diff erences in temperature and sunshine exposure by region

Table 10.2 illustrates the regional and within-year variation in temperature and sunshine

exposure in The Netherlands. For convenience we listed temperature and sunshine exposure

overall, and for three temperature stations separately: ‘de Kooy’, ‘de Bilt’, and ‘Maastricht’

located in the north, middle and south of The Netherlands, respectively. As expected,

temperatures are highest in summer (peaks in July / August, average 17.9 and 17.7 degrees

Celsius), and lowest in winter (nadirs in December / January, average 4.0 and 4.2 degrees

Celsius). Sunshine exposure is highest in May / June (average 7.2 and 7.3 hours per day),

and lowest in December / January (average 2.2 and 1.9 hours per day).

There is a difference of 1 degree Celsius between the northern (higher in January) and the

southern temperature stations (higher in July). In the north, the temperature extremes

appear less, with more sunshine, particularly in July with a difference of almost an hour

per day (7.4 - 6.6 = 0.8 hours).

Linear regressions

Table 10.3 lists the linear regressions with beta coefficients, their 95% confidence intervals

(CI), and the adjusted R2 values for each model. In model 0 all determinants have a

significant contribution to the model with the largest beta coefficients for primiparous

women (-169.8, CI -171.3;-168.4), children born at 37 weeks (-552.3, CI -555.5;-549.1) and 38

weeks (-324.1, CI -326.4;-321.9), and female sex (-145.8, CI -147.2;-144.3). In model 1, month

of birth was added which did not alter the beta coefficients of maternal and child factors.

180

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

Most months showed significant, yet small effects on birthweight with beta coefficients

ranging from 3.7 (September) to 6.7 (May) compared to January. No clear pattern in beta

coefficients of month of birth was observed. Temperature exposure measures were added

in model 2. A clear pattern in months emerged with a highly significant birthweight nadir

in summer: >18 grams lower birthweight in July (-18.4) and August (-18.5). Beta coefficients

for temperature exposure measures were small. For minimum temperature only exposures

during first, second and third trimester were significant with beta coefficients ranging

Table 10.2 Monthly differences of temperature and sunshine; overall and for three temperature stations separately: ‘de Kooy’, ‘de Bilt’, and ‘Maastricht’ located in the north, middle and south of The Netherlands, respectively

Mean day temperature

Minimum day temperature

Maximum day temperature

Mean day sunshine hours

Temperature station: all

MonthJanuary 4.2 -16.8 15.9 2.2February 4.2 -15.4 18.0 3.5March 6.2 -20.7 22.2 4.4April 9.6 -7.9 29.7 6.4May 13.6 -1.5 32.6 7.2June 16.1 0.9 34.7 7.3July 17.9 3.7 37.1 6.8August 17.7 4.4 37.8 6.0September 15.3 0.5 31.3 5.3October 11.2 -8.5 24.0 4.0November 7.4 -7.2 19.4 2.2December 4.0 -17.0 16.3 1.9

Temperature station: 'De Kooy'

MonthJanuary 4.8 -8.5 12.2 2.3July 17.5 8.4 30.2 7.4

Temperature station: 'De Bilt'

MonthJanuary 4.3 -10.6 14.1 2.1July 18.1 6.4 35.7 6.6

Temperature station: 'Maastricht'

MonthJanuary 3.9 -11.7 15.5 2.2July 18.5 7.2 36.3 6.6

181

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

Ta

ble

10

.3

Line

ar re

gres

sion

mod

els

with

regr

essi

on c

oeffi

cien

ts w

ith (9

5% c

onfid

ence

inte

rval

s)

(adj

uste

d R2 )

Mod

el 0

(0.2

1)M

odel

1 (0

.21)

Mod

el 2

(0.2

1)M

odel

3 (0

.21)

Beta

coe

ffi ci

ent

Beta

coe

ffi ci

ent

Beta

coe

ffi ci

ent

Beta

coe

ffi ci

ent

Cons

tant

3903

.9 (3

900.

9;39

07.0

)*39

00.0

(389

6.1;

3903

.8)*

*40

49.2

(403

5.7;

4062

.6)*

4019

.0 (4

004.

0;40

33.9

)*

Ma

tern

al

fac

tors

Parit

yPr

imip

arou

s-1

69.8

(-17

1.3;

-168

.4)*

-169

.8 (-

171.

3;-1

68.3

)*-1

69.9

(-17

1.4;

-168

.4)*

-169

.9 (-

171.

4;-1

68.4

)*M

ultip

arou

sRe

fere

nce

Refe

renc

eRe

fere

nce

Refe

renc

e

Mat

erna

l age

^2-0

.17

(-0.1

8;-0

.16)

*-0

.17

(-0.1

8;-0

.16)

*-0

.17

(-0.1

8;-0

.16)

*-0

.17

(-0.1

8;-0

.16)

*

Ethi

city

Wes

tern

Refe

renc

eRe

fere

nce

Refe

renc

eRe

fere

nce

Non

-Wes

tern

-90.

0 (-9

2.1;

-87.

9)*

-90.

0 (-9

2.1;

-87.

9)*

-91.

0 (-9

3.1;

-88.

9)*

-91.

4 (-9

3.5;

-89.

3)*

Soci

oeco

nom

ic s

tatu

s sc

ore

-17.

9 (-1

8.6;

-17.

1)*

-17.

9 (-1

8.6;

-17.

1)*

-18.

1 (-1

8.8;

-17.

3)*

-18.

0 (-1

8.7;

-17.

2)*

Ch

ild

fa

cto

rs

Ges

tatio

nal a

ge37

wee

ks-5

52.3

(-55

5.5;

-549

.1)*

-552

.3 (-

555.

5;-5

49.0

)*-5

52.2

(-55

5.4;

-549

.0)*

-552

.2 (-

555.

4;-5

49.0

)*38

wee

ks-3

24.1

(-32

6.4;

-321

.9)*

-324

.1 (-

326.

4;-3

21.8

)*-3

24.1

(-32

6.4;

-321

.8)*

-324

.1 (-

326.

4;-3

21.8

)*39

wee

ks-1

52.5

(-15

4.4;

-150

.5)*

-152

.5 (-

154.

4;-1

50.5

)*-1

52.4

(-15

4.4;

-150

.4)*

-152

.4 (-

154.

4;-1

50.5

)*40

wee

ksRe

fere

nce

Refe

renc

eRe

fere

nce

Refe

renc

e41

wee

ks12

7.7

(125

.6;1

29.8

)*12

7.7

(125

.6;1

29.8

)*12

7.6

(125

.5;1

29.8

)*12

7.6

(125

.5;1

29.7

)*≥4

2 w

eeks

218.

4 (2

15.0

;221

.8)*

218.

4 (2

15.0

;221

.8)*

218.

5 (2

15.1

;221

.8)*

218.

4 (2

15.1

;221

.8)*

Cong

enita

l ano

mal

ies

No

Refe

renc

eRe

fere

nce

Refe

renc

eRe

fere

nce

Yes

-74.

6 (-7

9.6;

-69.

5)*

-74.

6 (-7

9.6;

-69.

5)*

-74.

3 (-7

9.4;

-69.

3)*

-74.

3 (-7

9.3;

-69.

2)*

Tabl

e 10

.3 co

ntin

ues o

n ne

xt p

age.

182

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

Ta

ble

10

.3

Cont

inue

d

(adj

uste

d R2

)M

odel

0 (0

.21)

Mod

el 1

(0.2

1)M

odel

2 (0

.21)

Mod

el 3

(0.2

1)

Beta

coe

ffi ci

ent

Beta

coe

ffi ci

ent

Beta

coe

ffi ci

ent

Beta

coe

ffi ci

ent

Sex M

ale

Refe

renc

eRe

fere

nce

Refe

renc

eRe

fere

nce

Fem

ale

-145

.8 (-

147.

2;-1

44.3

)*-1

45.8

(-14

7.2;

-144

.3)*

-145

.7 (-

147.

2;-1

44.3

)*-1

45.7

(-14

7.2;

-144

.3)*

Cli

ma

tolo

gic

al

fac

tors

Day

of b

irth:

con

secu

tive

num

ber i

n st

udy

perio

d0.

010

(0.0

09;0

.010

)*0.

010

(0.0

09;0

.010

)*0.

010

(0.0

10;0

.011

)*0.

009

(0.0

09;0

.010

)*

Mon

th o

f birt

hN

ot in

equ

atio

nJa

nuar

y-

Febr

uary

-2.

4 (-1

.2;6

.0)

-1.3

(-5.

4;2.

8)-0

.3 (-

4.4;

3.8)

Mar

ch-

4.0

(0.5

;7.5

)***

-4.4

(-9.

3;0.

6)-2

.0 (-

7.1;

3.1)

Apr

il-

5.6

(2.0

;9.1

)**

-6.0

(-12

.0;-0

.1)*

**-1

.5 (-

7.7;

4.7)

May

-6.

7 (3

.2;1

0.2)

*-8

.3 (-

15.0

;-1.6

)***

-1.3

(-8.

4;5.

7)Ju

ne-

5.2

(1.7

;8.8

)**

-16.

1 (-2

3.3;

-8.9

)*-7

.5 (-

15.2

;0.2

)Ju

ly-

6.2

(2.8

;9.7

)*-1

8.4

(-25.

5;-1

1.2)

*-9

.3 (-

17.0

;-1.5

)***

Augu

st-

3.0

(-0.5

;6.5

)-1

8.5

(-25.

5;-1

1.4)

*-1

1.1

(-18.

7;-3

.6)*

*Se

ptem

ber

-3.

7 (0

.2;7

.2)*

**-1

5.1

(-21.

6;-8

.6)*

-10.

5 (-1

7.3;

-3.6

)**

Oct

ober

-4.

1 (0

.6;7

.6)*

**-6

.1 (-

11.6

;-0.6

)***

-3.5

(-9.

3;2.

3)N

ovem

ber

-5.

9 (2

.3;9

.4)*

*3.

5 (-1

.3;8

.3)

5.4

(0.4

;10.

4)**

*D

ecem

ber

-0.

4 (-3

.1;4

.0)

-0.4

(-4.

6;3.

8)0.

3 (-4

.0;4

.5)

183

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

Min

imum

tem

pera

ture

: win

dow

of e

xpos

ure

Not

in e

quat

ion

Not

in e

quat

ion

E1: p

eric

once

ptio

nal

--

0.3

(0.0

;0.6

)0.

2 (-0

.1;0

.6)

E2: fi

rst t

rimes

ter

--

-0.6

(-1.

1;-0

.2)*

*-0

.3 (-

0.8;

0.1)

E3: s

econ

d tr

imes

ter

--

0.6

(0.3

;1.0

)**

0.4

(0.0

;0.8

)***

E4: t

hird

trim

este

r-

-1.

5 (1

.3;1

.8)*

1.5

(1.2

;1.7

)*E5

: day

of b

irth

--

0.1

(-0.2

;0.4

)0.

1 (-0

.2;0

.4)

Max

imum

tem

pera

ture

: win

dow

of e

xpos

ure

Not

in e

quat

ion

Not

in e

quat

ion

E1: p

eric

once

ptio

nal

--

-0.6

(-0.

9;-0

.3)*

-0.4

(-0.

8;-0

.1)*

*E2

: fi rs

t trim

este

r-

--1

.6 (-

1.9;

-1.2

)*-2

.1 (-

2.4;

-1.7

)*E3

: sec

ond

trim

este

r-

--2

.2 (-

2.5;

-1.9

)*-2

.4 (-

2.8;

-2.0

)*E4

: thi

rd tr

imes

ter

--

-2.0

(-2.

3;-1

.7)*

-2.1

(-2.

5;-1

.8)*

E5: d

ay o

f birt

h-

--0

.4 (-

0.7;

-0.2

)**

-0.4

(-0.

6;-0

.1)*

*

Cum

ulat

ive

suns

hine

in h

ours

: win

dow

of e

xpos

ure

Not

in e

quat

ion

Not

in e

quat

ion

Not

in e

quat

ion

E2: fi

rst t

rimes

ter

--

-0.

052

(0.0

39;0

.065

)*E3

: sec

ond

trim

este

r-

--

0.02

5 (0

.011

;0.0

38)*

*E4

: thi

rd tr

imes

ter

--

-0.

023

(0.0

10;0

.036

)**

* p<

0.00

01, *

* p>

0.00

01 a

nd p

<0.0

1, *

** p

>0.0

1 an

d p<

0.05

.

184

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

from -0.6 to 1.5. For maximum temperature all exposure windows had significant beta

coefficients ranging from -0.4 (day of birth) to -2.2 (second trimester). After adding sunshine

exposure to model 3, the summer nadir remained significant, but with moderation of beta

coefficients to -9.3 for July and -11.1 for August. Sunshine exposure demonstrated small,

yet significant effects with beta coefficients ranging from 0.023 to 0.052. As opposed to

Figure 10.3 Regional birthweight differences (per 2-digit zip code) attributable to difference in exposure to temperature extremes and sunshine (adjusted for maternal factors, child factors, and day and month of birth).

185

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

the emerging pattern of seasonality in model 2, there was no difference in adjusted R2

(0.21) between the models, implying combined (interacting) effects of temperature and

sunshine with seasonality.

Regional birthweight diff erences

The regional birthweight differences range from -11 to +25 grams due to the selected

climatological exposures, with moderation of these effects (i.e., higher birthweights) in

coastal areas (figure 10.3).

DISCUSSIONBirthweight is associated with the combined effect of season and exposure to temperature

extremes and cumulative exposure to sunshine, in particular during critical reproductive

exposure windows in pregnancy. We demonstrated small, yet consistent effects on top of

known maternal and child factors. The effects of seasonality alone were initially modest

(table 10.3, model 1) but most likely hided interacting effects of temperature and sunshine,

illustrated by the clear pattern in seasonality after inclusion of temperature and sunshine

in the model (table 10.3, model 2 and 3). The result was an almost 20 gram significantly

decreased birthweight in summer compared to January. We found minimum temperature

exposure in the second and third trimester to be associated with higher birthweight,

maximum temperature exposure for all exposure windows to be associated with lower

birthweight, and cumulative sunshine exposure to be associated with higher birthweight.

On the population level, there are significant regional differences in birthweight, attributable

to the selected climatological exposure measures, ranging from -11 to +25 grams (relative

to the observed birthweight). The detrimental effect of maximum temperature in particular,

appears to be moderated in coastal areas where there are less temperature extremes and

more sunshine hours. The superposition of the observed effects with differential timing of

exposure explains the hitherto unexplained or contradictory patterns in previous papers.6

Other studies

Previous studies on the relationship between climate and birthweight are heterogeneous

in design and outcome. 6-10 In an Australian large cohort study in the Perth region, Pereira et

al. investigated the effect of seasonal variation, temperature and sunshine on birthweight,

adjusted for sociodemographic, biological, and environmental exposures (e.g., air pollutants,

186

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

temperature and sunshine).7 Effects were studied separately by trimester. Analogous to

our findings, they report a 2% significantly increased risk of SGA by higher temperatures

sustained over pregnancy; this risk for lower birthweight was particularly observed for

exposition during the third trimester. The main differences compared to our study include

the use of only maximum temperature, exposure data from just two temperature stations,

and the inclusion of births from 33 weeks of gestational age, leading to incomplete exposure

windows for the third trimester.7

Although an inverse relation of heat stress on birthweight has been observed in several

studies7,9,15, the opposite (i.e., heat stress causing increased birthweight) has also been

demonstrated.8,9 Probably due to heterogeneity in study design, most of these studies

did not find consistent directions of effect measures for all exposure windows, e.g., one

study found a negative effect of first trimester temperature exposure on birthweight

(beta coefficient -5.4, 95% CI -7.9;-2.9), and a positive effect of third trimester temperature

exposure on birthweight (beta coefficient 1.3, 95% CI 0.5-2.1).9

Animal studies provide a biological explanation for part of our findings as they show

heat stress during pregnancy to be associated with reduced placental weight, decreased

uterine and umbilical blood flow and consequent reduction in offspring birthweight.16

Galan et al. exposed pregnant ewes to heat stress during pregnancy for 55 days (mean

birthweight 1,841 grams) and 80 days (mean birthweight 882 grams).17 They observed a

significant reduction in fetal and placental weights compared with a control group (mean

birthweight 3,391 grams).17 In terms of the ‘fetal origins of adult disease’ hypothesis5, our

findings, combined with findings from these animal studies may demonstrate some form

of adaptation mechanism regarding the expected outdoor climate after birth.

Possible mechanisms

Several theories exist on the background of our observed results. According to one, fetal

growth is influenced by the rate of uteroplacental blood flow. Extremes of temperature, in

particular heat, are known to affect human blood flow with excess cardiovascular deaths

occurring in heat waves.11 It is therefore plausible that maternal blood flow and, hence,

fetal nutrition will be affected by these extremes in different stages of pregnancy. It is also

possible that outdoor temperature during pregnancy or seasonality is linked to maternal

behaviours including smoking, diet and physical activity. In particular smoking may have

contributed to the summer decrease in birthweight as it has a profound negative effect on

birthweight, and tobacco consumption in general has been shown to peak in summer.18,19

187

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

Inherent to sunshine, vitamin D has also been mentioned as a possible causal factor. Its

production depends primarily on the action of sunlight on the skin, and therefore, it is

strongly associated with the duration of sunshine.12,20 A 2010 Dutch study found women

with deficient vitamin D levels in early pregnancy (median 13 weeks’ gestation) to be at

increased risk for an infant with lower birthweight (-114.4 gram, 95% CI -151.2;-77.6) and

SGA.20 Vitamin D deficiency is more prevalent during the winter months. Women with the

end of first trimester in winter will deliver during late spring and summer, and, according

to this hypothesis, will have lower birthweight children as demonstrated in our results.

Strengths and limitations

An important strength of our study is the usage of a validated national database (The

Netherlands Perinatal Registry) with an almost complete coverage of all pregnancies over

a long period of time (2000-2008), and detailed climatological data from a nationwide

network of 29 temperature stations, increasing the precision and generalisability of our

results. Additionally, in comparison with previous studies, our study model included an

individual exposition model covering the whole duration of pregnancy including the

periconceptional period.3,6-10 Other strengths include the use of sunshine exposure, as most

studies do not include this in their model, and the adjustment for socioeconomic status.

The latter is important because in current times it is relatively easy to keep a home and a

person in a temperate ambient, but this is less likely the case if socioeconomic status is low

and housing quality is poor.

One of the study limitations, similar to most previous studies, is that outdoor temperature

does not reflect the true exposition as a significant amount of time is spent indoors,

especially at time of extreme temperatures. Another potential source of error is a person’s

mobility as climatological data linked to a person’s zip code of residence may compromise

the exposure estimate. Inaccuracy resulting from the first underestimates the individual

exposure to temperature extremes with underestimation of the true peak temperature

effect. The second inaccuracy may yield an under- or overestimation on the individual

level without obvious systematic effect in general. Despite this error and the generally

temperate Dutch climate, a consistent effect remains. A last limitation is the lack of data

on maternal lifestyle factors, in particular smoking. However, by excluding preterm births

and taking into account socioeconomic status, both strongly related to maternal smoking,

we attempted to minimise the effects of this limitation.

188

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

Implications

Climatological factors impact birthweight on the individual level, the effect of which

emerges as regional differences in birthweight on the population level. For healthy babies

born at term the demonstrated differences appear small and effects on health later in life

are unlikely. More detrimental effects may emerge for vulnerable subgroups (e.g., women

carrying growth restricted babies) implying future research in these groups. The birthweight

effects on the population level may be also more sizeable in countries where the population

is truly exposed to extremes.

The World Health Organization recommends an indoor temperature of 18 degrees Celsius,

or 21 degrees Celsius for vulnerable people21; as it is impossible to change the climate,

improving housing quality to moderate exposure to extremes in temperature appears a

realistic implication for vulnerable subgroups.

Conclusion

Birthweight is associated with the combined effect of season and exposure to temperature

extremes and cumulative exposure to sunshine, in particular during critical reproductive

exposure windows in pregnancy. The effect of seasonality appears to be mediated by

interacting effects of temperature and sunshine, resulting in an almost average 20 gram

significantly decreased birthweight in summer compared to January. On the population

level, significant regional differences in birthweight exist, most likely attributable to

particularly maximum temperatures, with moderation of the effect in coastal areas. For

healthy babies born at term these differences in birthweight appear small; however, more

severe detrimental effects may emerge for vulnerable subgroups, e.g., women carrying

growth restricted babies.

189

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

1. Hippocrates. Airs, waters, places. In: Lloyd G, ed. Hippocratic writings. London: Penguin books; 1978:148-169.

2. Bhaskaran K, Armstrong B, Hajat S, Haines A, Wilkinson P, Smeeth L. Heat and risk of myocardial infarction: hourly level case-crossover analysis of MINAP database. BMJ. 2012;345:e8050.

3. Seretakis D, Lagiou P, Lipworth L, Signorello LB, Rothman KJ, Trichopoulos D. Changing seasonality of mortality from coronary heart disease. JAMA. 1997;278:1012-1014.

4. Lambert GW, Reid C, Kaye DM, Jennings GL, Esler MD. Effect of sunlight and season on serotonin turnover in the brain. Lancet. 2002;360:1840-1842.

5. Barker DJ. Adult consequences of fetal growth restriction. Clin Obstet Gynecol. 2006;49:270-283.

6. Strand LB, Barnett AG, Tong S. The influence of season and ambient temperature on birth out-comes: a review of the epidemiological literature. Environ Res. 2011;111:451-462.

7. Pereira G, Cook A, Haggar F, Bower C, Nassar N. Seasonal variation in fetal growth: accounting for sociodemographic, biological, and environmental exposures. Am J Obstet Gynecol. 2012;206:74 e71-77.

8. Murray LJ, O’Reilly DP, Betts N, Patterson CC, Davey Smith G, Evans AE. Season and outdoor ambient temperature: effects on birth weight. Obstet Gynecol. 2000;96:689-695.

9. Lawlor DA, Leon DA, Davey Smith G. The associa-tion of ambient outdoor temperature throughout pregnancy and offspring birthweight: findings from the Aberdeen Children of the 1950s cohort. BJOG. 2005;112:647-657.

10. Torche F, Corvalan A. Seasonality of birth weight in Chile: environmental and socioeconomic factors. Ann Epidemiol. 2010;20:818-826.

11. Huynen MMTE, Martens P, Schram D, Weijenberg MP, Kunst AE. The impact of heat waves and cold spells on mortality rates in the Dutch population. Environ Health Persp. 2001;109:463-470.

12. Urrutia RP, Thorp JM. Vitamin D in pregnancy: current concepts. Curr Opin Obstet Gynecol. 24: 57-64.

13. Wilcox AJ. Fertility and pregnancy: an epidemio-logic perspective. New York: Oxford University Press; 2010.

14. Perinatal care in The Netherlands 2008 (in Dutch: Perinatale zorg in Nederland 2008). Utrecht: The Netherlands Perinatal Registry; 2011.

15. Wells JC, Cole TJ. Birth weight and environmental heat load: a between-population analysis. Am J Phys Anthropol. 2002;119:276-282.

16. Wells JC. Thermal environment and human birth weight. J Theor Biol. 2002;214:413-425.

17. Galan HL, Hussey MJ, Barbera A, et al. Relationship of fetal growth to duration of heat stress in an ovine model of placental insufficiency. Am J Obstet Gynecol. 1999;180:1278-1282.

18. Andres RL, Day MC. Perinatal complications associated with maternal tobacco use. Semin Neonatol. 2000;5:231-241.

19. Warren CW, Erguder T, Lee J, et al. Effect of policy changes on cigarette sales: the case of Turkey. Eur J Public Health. 2012;22:712-716.

20. Leffelaar ER, Vrijkotte TG, van Eijsden M. Maternal early pregnancy vitamin D status in relation to fetal and neonatal growth: results of the multi-ethnic Amsterdam Born Children and their Development cohort. Br J Nutr.;104:108-117.

21. Health impact of low indoor temperatures: Report on a WHO meeting. Copenhagen: World Health Organization; 1985.

REFERENCES

190

EXTREMES IN

OU

TDO

OR TEM

PERATURE A

ND

BIRTHW

EIGH

T

10

11General discussion

192

GEN

ERAL D

ISCUSSIO

N

11

AIM OF THIS THESISThe aim of this thesis was to investigate the main contributing factors in adverse perinatal

outcomes on the three geographic levels (international, national, regional). The role of

patient, environmental, and healthcare related factors was studied, in particular their

relative importance. Such information may guide to select and to prioritise among policies

to decrease gaps in perinatal health.

Our results are presented in two parts. Part I contains three studies from the regional ‘Ready

for a Baby’ programme; part II contains six studies on the national level, either ordered

by ‘ZonMw’ (The Netherlands Organisation for Health Research and Development), ‘SAZ’

hospitals (a co-operative institution of about 40 small general hospitals in The Netherlands),

or the Ministry of Health, Welfare and Sport. All studies draw on data from The Netherlands

Perinatal Registry.

PRINCIPAL FINDINGS

Part I

We demonstrated that large differences exist in absolute perinatal mortality and morbidity

rates between neighbourhoods within the city of Rotterdam (chapter 2). The magnitude

of these inner-city inequalities in perinatal health had not been previously described in

developed countries. The inequalities were somewhat reduced by accounting for socio-

demographic population differences (‘standardisation’ for maternal age, parity, ethnicity,

neighbourhood social quality) between the Rotterdam boroughs, yet substantial differences

remained (chapter 3). The municipal report on perinatal health as described in chapter

3 shows how such an analysis allows to differentiate between patterns of possible causes

per borough, implying the necessity of different policy measures per borough. We believe

that the mechanisms in deprived neighbourhoods work through the accumulation of

heterogeneous risks (maternal, child, organisational, environmental). These risks are more

abundant in these neighbourhoods and underlie the generally high level of adverse

pregnancy outcomes.1

Another important factor influencing urban perinatal health is social deprivation (chapter

4). Deprivation points both to the above described higher risk level, and the lack of

personal and family resources to counter ill-health. We demonstrated differential effects

of social deprivation on Western and non-Western women in the city of Rotterdam. Better

193

GEN

ERAL D

ISCUSSIO

N

11

neighbourhood social quality (expressed in the Social Index) was clearly associated with

better perinatal outcomes for Western women. This trend was absent for non-Western

women as they showed small changes in prevalence of adverse pregnancy outcome with

better neighbourhood social quality. Policy recommendations in this context may refer

to (1) risk management and intensified care, and (2) empowerment, improving resources

and care access. For Western women, the latter is more important as our results show

that improved social quality is associated with perinatal outcome improvement in this

group. While this may also benefit non-Western women, tailored change requires better

understanding of the underlying patterns first. Suggested explanations for non-Western

‘migrant’ groups include the presence of ‘protective’ effects through knowledge systems

(‘resource availability’) or intrinsic resilience (‘high risk level’).2-4 The suggestion of knowledge

systems follows from the so-called ‘ethnic density’ hypothesis which suggests the risk of poor

health outcomes for a minority individual to be inversely related to the density of his or her

racial/ethnic group in the local community.5-8 The presumed mechanisms include health

protection through increased participation in social networks or knowledge systems, and a

more extensive repertoire of positive coping behaviours.5-9 Additional etiological research

in this multi-ethnic urban setting is mandatory. Possibly, risk assessment tools and social

measures such as the Social Index capture features through a ‘Western perspective’, e.g.,

rather than household income greater family income may be a more important measure

for non-Western groups.

Part II

The population attributable risks (PAR10) of maternal, child and organisational risk factors for

perinatal mortality are demonstrated in chapter 5. Adapting a new multivariable method

to compute PARs, the results showed to vary according to methodological choices. For

example, a PAR of maternal factors can be estimated with and without taking into account

the PAR of child and organisational factors. Maternal factors, long put forward as a crucial

determinant for the national perinatal mortality rate11-13, appeared to attribute to only 1.4%

to 13.1% of perinatal mortality in a multivariate context. PARs of child and organisational

factors were 58.7-74.0% and 7.3-34.3%, respectively, where the range reflects different

priority (for that factor) in the adjustment procedure. Focusing on the role of hospital organi-

sation factors, the PAR results suggest that a change towards the theoretical organisation

optimum, with an otherwise unchanged pregnant population, provides a 34.3% decrease

of in-hospital perinatal mortality.

194

GEN

ERAL D

ISCUSSIO

N

11

The highest PAR for a single risk factor was observed for gestational age (72.2%), i.e., if all

women would deliver at 40 weeks of gestational age - all other things equal - , the perinatal

mortality is expected to be reduced with 72.2%. Moreover, being born at 37-37.6 weeks

of gestation (considered at term) reflects a 200% increased risk for perinatal mortality

(compared to 40 weeks), which questions the definition of ‘term pregnancy’ and suggests

a change of the lower threshold from 37 weeks to 38 weeks of gestation.14,15

Results from the PAR study are in support of policy measures aimed at (1) improving organi-

sational factors in obstetric healthcare16-19, and (2) improved prevention of preterm birth.

Next to targeting specific risk factors, also regional policy measures may be implemented

to improve perinatal health. Chapter 6 reported on epidemiological methods to select

priority regions in which to study the implementation of intensified preconception care

and uniformal antenatal risk selection for the ‘Healthy Pregnancy 4 All’ project. Crude as

well as standardised rates are used in this selection process. Depending on the goal of the

intervention, i.e., improvement of preconception care or antenatal risk selection, different

indicators are needed to prioritise among regions. These indicators are related to perinatal

mortality, morbidity and healthcare factors. Both strategies of selection point out the four

largest cities (‘G4’, i.e., Amsterdam, Rotterdam, The Hague, Utrecht) as a priority region,

thus, illustrating the multifactorial effects on and the complexity of urban perinatal health.

Chapters 7, 8 and 9 address the unique Dutch system of obstetric care. This system is

characterised by three risk-based levels of care.20,21 Primary care for assumed low risk

pregnancies is provided by independently practicing community midwives. Assumed low

risk pregnant women can either opt for a home birth or a short-stay hospital birth under

supervision of a community midwife.20,21 Secondary/tertiary care for assumed high risk

pregnancies is provided by obstetricians in hospitals. Whenever risk factors of adverse

perinatal or maternal outcome are present before pregnancy or arise during pregnancy

or parturition, women shift from low risk to high risk and are referred to secondary care

or from secondary to tertiary care, also during parturition. This ongoing risk assessment

during pregnancy and during parturition is called ‘risk selection’, an essential quality of

care indicator of the Dutch obstetric care system.20-22 Risk selection in the Netherlands is

the exclusive responsibility of primary care community midwives.22 In formal terms, the

aim of risk selection is to identify and refer high risk pregnancies from an undifferentiated

group at onset of pregnancy, in order to obtain a true low risk group of pregnant women

(expressed as high negative predictive value of risk selection) for delivery in the primary

care setting.20,21

195

GEN

ERAL D

ISCUSSIO

N

11

Indications for referral and definitions of high and low risk pregnancy are listed in the ‘List of

Obstetric Indications’ (in Dutch: Verloskundige Indicatie Lijst).20,22 These indications, however,

are not fully registered as such in The Netherlands Perinatal Registry, and not always defined

unequivocally. For our analysis of risk selection effectiveness we therefore defined a high

risk pregnancy based on the presence of a ‘Big4’ morbidity.23

Our study showed that current risk selection does not result in a true low risk population

in primary care at the end of pregnancy. Furthermore, a considerable part of high risk

women is referred during parturition (chapter 7). Several mechanisms may be responsible

for suboptimal risk selection. First, primary care community midwives are oriented towards

low risk pregnancies. This ‘risk experience’ may cause high risk women not being detected

as such in primary care. By a similar mechanism low risk women who might deliver under

a midwife’s supervision may not always be detected in secondary/tertiary care. Second,

the availability of tools to detect high risk pregnancies in primary care may also play a role,

e.g., standardised checklist-based risk screen methods in the first trimester or ultrasound

examination in the second and third trimester. Standardised checklist-based antenatal risk

selection is currently evaluated in a national public health programme: ‘Health Pregnancy 4

All’ (HP4All).24 The instrument being used, the ‘R4U’ (Rotterdam Reproductive Risk Reduction),

screens on medical as well as social risk factors for adverse perinatal outcome.24

Financial incentives are a third mechanism which may influence risk selection effectiveness.

Assigning a high risk level implies referral, which directly affects professional income.

For example, midwives receive the full fee for delivery care, independent of the duration

of labour. If the patient is referred just after (artificial) rupture of membranes, this does

not affect the fee. Similarly, no surplus is provided in time consuming deliveries. This

incentive promotes a policy in which pregnant women stay under care of the primary care

community midwife until just after the start of delivery. Similarly, gynaecologists receive a

fee independent from the stage of delivery. The two aforementioned incentives together

may be responsible for high referral rates during parturition.

We believe that a successful effort in increasing risk selection effectiveness depends on

optimal collaboration between midwives and obstetricians with a joint responsibility for

the determination of a woman’s risk status, i.e., ‘shared care’. Shared obstetric care has

already been implemented in some form in other Western countries such as Australia and

the United Kingdom.25-27 One study demonstrated a 27% increase in the detection rate of

intrauterine growth restriction for women receiving shared obstetric care as opposed to

conventional obstetric care.28

196

GEN

ERAL D

ISCUSSIO

N

11

Chapter 8 addresses the safety of home birth. Under routine conditions, home birth is

generally not associated with increased intrapartum and early neonatal mortality. The

mortality risk of home birth compared to planned hospital birth in low risk pregnancies is

non-significantly increased (OR 1.05, 95% CI 0.91–1.21). In subgroups, additional risk cannot

be excluded. Moreover, a recent report, which presents an overview of evidence concerning

safety of home birth, advocates hospital based delivery for all women. The evidence in their

view supports professional responsibility to be more important than a consumer’s choice,

i.e., the home birth option for low risk women.29

In this thesis we frame the policy choice involving the home birth option along two key

questions. The first is whether risk selection is adequate, i.e., is the group starting delivery

under a midwife’s supervision truly a low risk group of women, who consequently can

be offered the choice to deliver at home? A 6% prevalence of perinatal morbidities in

home births - even after referral of high risks during parturition and pregnancy - suggests

a negative reply to this answer (chapter 7). The second question refers to the outcome

difference in the home environment vs. in the hospital environment in unanticipated

perinatal morbidities. We assume that neonates with morbidities are likely to benefit from

being in the hospital setting at onset, as this decreases delay in specialised care provision.30,31

In-hospital midwife-led birth centres provide a possible alternative to provide a home-like

environment during delivery without delay if referral to the hospital is deemed necessary.32

While home birth perhaps may be considered for a group of selected low risk multiparous

women, one can argue that all primiparous women should deliver in an in-hospital

environment, either under supervision of the midwife (birth centre) or a gynaecologist.29,33

The study in chapter 9 was set up because of current policy proposals in favour of cen-

tralisation acute obstetric care. The goal of centralisation is to ensure 7*24h availability of

‘qualified professionals’ (midwives, gynaecologists, paediatricians, anaesthesiologists, and

operating theatre staff; ‘qualified’ in terms of seniority) within 15 minutes.16,34

Centralisation is also under consideration in other countries.17-19 Recently, the Royal Col-

lege of Obstetricians and Gynaecologists (RCOG) stated in their ‘Working Party Report’ it

is adamant that obstetric care delivery has qualified specialists available at all times, 24

hours a day, 7 days a week as more than half of all births, after all, take place ‘out of hours’.17

However, also taking into account the possible drawbacks of centralisation they caution

this recommendation by stating ‘localised where possible, centralised where necessary’.17

The main reasons mentioned in the literature in favour of centralisation are: (1) better

care and access to rapid intervention during delivery, resuscitation of the newborn, and

197

GEN

ERAL D

ISCUSSIO

N

11

identification and management of newborn infants with unexpected morbidities (e.g.,

congenital anomalies) in large and higher level hospitals.35 However, this ‘bigger-is-better’

mantra is mainly based on theoretical reasoning, and empirical evidence from centralisation

studies in the context of (complex) surgical procedures.36 Moreover, in the context of

acute obstetric care, evidence-based data are scarce, and most studies on this subject are

ecological or of a descriptive nature, heterogeneous in design and have contradictory

results.18,19,35,37,38 One important argument against centralisation is increased travel time

to hospital, with inherent increased risk for adverse perinatal outcome (especially for high

risk women) or out-of-hospital delivery.39 Furthermore, as around 20% of births in The

Netherlands still occur at home under supervision of a midwife40, increased travel time to

hospital will compromise some low risk women in their choice to opt for a birth at home.

Finally, centralisation of acute obstetric care fits into a broader change of Dutch hospitals

towards merging, strongly advocated by health insurance companies who prefer larger

institutions for management reasons.

We provided an empirical framework to study the effects of centralisation of acute obstetric

care (chapter 9). Two hypothetical centralisation scenarios are defined, both representing

fairly small changes affecting 6-7% of patients: (1) closure of the 10 smallest hospitals,

and (2) closure of the 10 smallest hospitals, but avoiding adjacent closures. As only 10

small hospitals were hypothetically closed we assumed no system effects. Our results

demonstrated that closing the small hospitals (scenario 1 and 2) does not necessarily result

in distinct better outcomes on the population level. Outcome strongly depended on the

organisational features of the hospital receiving the patients from the closed hospitals.41

Multiple factors have to be taken into account regarding centralisation of acute obstetric

care, e.g., the organisational features of the closing hospital, the characteristics of the

women ‘shifting’ to the next nearest hospital (are they low risk or high risk), the quality of

risk selection (are there many ‘unexpected’ high risk women who need to be transferred);

thus, also additional travel time is an essential feature. Moreover, short term and long term

economic aspects of centralisation and ‘side’ effects on paediatric and anaesthesiology

services also need to be evaluated.

In line with the results from chapter 5, showing a 30% decrease in perinatal mortality to

be expected from optimising hospital organisational factors alone, we think centralisation,

on average, to be effective in improving perinatal health. However, centralisation should

be carried out bearing (organisational) information on the specific hospitals in mind. In

particular rural regions may not benefit at all from centralisation. We therefore support the

RCOG’s statement ‘localised where possible, centralised where necessary’.17

198

GEN

ERAL D

ISCUSSIO

N

11

In chapter 10 we studied the effect of climatological factors, i.e., seasonality, extremes

in temperature and cumulative sunshine exposure, on birthweight. For this purpose we

specified an elaborate exposition model taking into account the whole course of pregnancy

from conception to birth. Our aim was to identify critical phases and the type of exposure

to demonstrate the strongest effects on birthweight.

We found all measures of exposure to be significantly associated with birthweight, where

the modest effects of season alone hided interacting effects of temperature and sunshine.

In a multivariable context, we found minimum temperature exposure in the second and

third trimester to be associated with higher birthweight, maximum temperature exposure

for all exposure windows to be associated with lower birthweight, and cumulative sunshine

exposure to be associated with higher birthweight. The greatest single effect was observed

for seasonality, in particular after correction for temperature and sunshine effects. Unlike

most previous studies42, we demonstrated a regular seasonal pattern after this correction,

resulting in a spring/summer nadir with an almost 20 gram significantly decreased

birthweight in summer. Our results partially correspond with results from animal studies

in which maximum temperature exposures are consistently significantly associated with

lower birthweight.43,44

On the population level and combining all seasons, we demonstrated significant regional

differences in birthweight, attributable to the studied measures of minimum and maximum

temperature, and cumulative sunshine exposure. Maximum temperatures appear to

attribute most to these regional effects, with moderation in coastal areas. For healthy babies

born at term these differences appear small; however, more severe detrimental effects may

emerge for vulnerable subgroups, e.g., women carrying growth restricted babies. Regarding

the effect of climatologic factors on birthweight, the population level is too small to expect

a contribution to the national problem of increased perinatal mortality.

METHODOLOGICAL ISSUES

The Netherlands Perinatal Registry

In all studies described in this thesis data from The Netherlands Perinatal Registry are used.

This medical registry contains complete population-based information of >97% of all

pregnancies in The Netherlands.40 Source data are routinely collected by 94% of midwives,

99% of gynaecologists and 68% of paediatricians including 100% of Neonatal Intensive Care

Unit paediatricians.40 (See website for detailed description: www.perinatreg.nl).

199

GEN

ERAL D

ISCUSSIO

N

11

In general, medical registries such as The Netherlands Perinatal Registry try to capture

particular actions in healthcare systems (e.g., admissions, billing, drug prescriptions), while

research registries usually try to capture details on one specific disease or research question.

The use of medical registries for research purposes may be challenged by the limited amount

and the quality of the information. Quality concerns include the classification of the health

outcome, the limited amount of detailed clinical information (e.g., existing health status

of the patient), the limited amount of information on events occurring before the health

event (e.g., information on past exposures) and information on events after the health

event (e.g., follow-up information).45

While useful for the studies described in this thesis, The Netherlands Perinatal Registry’s

quality is challenged in different ways. Health events and disease states are not always

clearly classified, e.g., presence of pre-eclampsia. Its quality is also challenged by the limited

amount of information before and after birth (the health event), on former pregnancies,

on risk factors (such as smoking, educational level, etc.), on process information of hospital

admission and referrals (e.g., indication for referral and treatment in such cases). Secondly,

about 70% of the paediatricians and 100% of NICU paediatricians participate. While the

percentage of participating paediatricians is increasing, partial and selective participation

challenges completeness of short term neonatal outcome. Thirdly, the registry does

not contain long term follow up outcome of newborns. This information is registered

separately by, e.g., care providers in youth healthcare. While these shortcomings may be

true in general, our studies primarily suffer from the unavailability of detailed information

on former pregnancies and risk factors, since these may lead to residual confounding. To

some extent they also suffer from missing data on referrals and hospital admissions, since

these may provide insight in the possible underlying mechanisms of adverse outcome. Most

importantly, the legal environment and user costs should be evaluated to ensure access

to and continuing improvement of The Netherlands Perinatal Registry as our studies show

the value of this registry for our research purposes.

Adjustment techniques

In the observational context of the studies presented in this thesis different adjustment

techniques are indispensible. We used multivariable regression techniques and (direct)

standardisation techniques.46-49 Advantages of direct and indirect standardisation include

computational simplicity, and relatively fewer statistical assumptions. Standardisation can

be regarded as the preferred adjustment technique if one is more interested in the overall

200

GEN

ERAL D

ISCUSSIO

N

11

effect instead of the influence of separate determinants.46-48 Moreover, standardisation

requires a researcher’s selection of adjustment variables prior to outcome analysis, which

forces thought and plausible rationale from the researcher. In our research, standardised

rates provided useful summary measures, especially when outcomes are rare and specific

rates display wide random variability. However, any summary measure can hide patterns

that might have important public health implications. Standardisation rates put more

emphasis on less prevalent groups.46-48

Adjustment by regression (forced entry of adjusters), is more convenient for statistical tests

for interactions and group differences (the individual effect of different determinants).

Therefore, when many determinants are present, adjustment by means of multivariable

regression analysis is more convenient.47-49 However, when applying this technique more

statistical assumptions have to be made (particularly when continuous variables are used),

such as linear correlation of adjustment variables with outcome effects, equal effects

for everyone, and all combinations of parameters being possible (including biological

implausible combinations).47-49 Moreover, in regression adjustment variables are less

restricted as they can be, e.g., nominal, ordinal or interval variables; adjustment variables in

standardisation, however, require strata to be created. Adjusted rates in regression are based

on the most prevalent groups. Also, regression can either require a researcher’s selection of

adjustment variables (forced entry) or this can be determined statistically (stepwise) which

may compromise plausibility.

In chapters 4, 5, and 9 we use multilevel modelling to adjust for clustering of individuals

within their neighbourhoods (chapter 4) or hospitals (chapters 5, 9). The advantage of this

approach is that it allows for the incorporation of both individual-level and neighbourhood-

level or hospital-level characteristics. Moreover, it increases the validity of the effect sizes

compared to conventional logistic regression analysis.50 In chapter 9 we observed great

variation in random effects per hospital suggesting differences in hospital performance in

terms of perinatal mortality.

Due to lacking information in The Netherlands Perinatal Registry, in particular individual-

level socioeconomic status, we, however, could not fully benefit from the theoretical

advantages of multilevel modelling over conventional logistic regression.

‘Big4’ morbidities

From detailed analysis of the complete perinatal dataset of The Netherlands Perinatal

Registry, it appeared that the presence of a so-called ‘Big4’ morbidity precedes perinatal

201

GEN

ERAL D

ISCUSSIO

N

11

mortality in 85% of cases.23 These ‘Big4’ morbidities are: congenital anomalies, preterm

birth, small for gestational age, and low Apgar score. Risk status of pregnancies could be

insufficiently established due to lacking data in The Netherlands Perinatal Registry, especially

due to lacking registration of referral indications. Due to their attribution to perinatal

mortality, we used the presence of one or more ‘Big4’ morbidities as a proxy measure of

high risk pregnancy in several studies in this thesis.

A limitation from this proxy measure may refer to ‘preventability’. From this retrospective

dataset no conclusions can be drawn on the preventability of ‘Big4’ morbidities in primary

care would they have been referred to secondary/tertiary care. However, congenital

anomalies and SGA are the most predictable in the antenatal phase. Most congenital

anomalies are now detected by routine ultrasound (introduced in 2007) at 20 weeks of

gestational age. Furthermore, there is general consensus on (improving) detection of SGA

which holds an increased risk for adverse perinatal outcome.51

Most important, however, is that, again, both general consensus and the ‘List of Obstetric

Interventions’ agree that a neonate with a ‘Big4’ morbidity is better off in a hospital setting

under the care of an obstetrician/paediatrician.52 Thus, the more important issue concerns

the adequate level of care for a high risk pregnancy (optimal risk selection) and not the

preventability of ‘Big4’. Recent Dutch reports have expressed concerns on the effectiveness

of risk selection in Dutch primary obstetric care.23,30,31 Results from these studies emphasise

that the level of healthcare provision is inadequate for a proportion of supposedly low

risk pregnant women at the onset of labour. Whether the delay in referral is related to late

diagnosis (no continuous fetal heart rate monitoring during parturition in primary care),

transport to hospital or assessment (‘primary care is supposedly low risk’), is yet unclear

and needs to be additionally studied.30,31

RECOMMENDATIONSFrom the above discussion the following recommendations can be given.

Recommendations Part I

• Within the city of Rotterdam, neighbourhood and borough-specific policies are mandatory

on tackling the high degree of perinatal health inequalities between neighbourhoods

and boroughs. These recommendations most likely will extend to other large cities.

202

GEN

ERAL D

ISCUSSIO

N

11

• In particular Western women in socially deprived neighbourhoods are entitled to

intensified antenatal care. This approach could include intensified prevention from

existing health promoting programs in combination with targeted social welfare. Also,

risk assessment should be modernised, regarding both the tools used for antenatal risk

screening and the approach of risks detected (‘shared care’). To enhance sustainability of

the joint responsibility of both obstetric and social caregivers in high risk pregnancies, a

combined financial remuneration system with adequate incentives should be advocated.

• Additional research is needed on (1) the differential effects of deprivation on Western

and non-Western pregnant women, (2) the risk accumulation phenomenon, and (3)

the consequences of adverse pregnancy outcomes in childhood. Concerning (1), the

sources of differential effects are important for tailored improvement programmes.

Concerning (2), the generalisability of the risk accumulation phenomenon is important

to support national antenatal screening instruments like the ‘R4U’ checklist. Concerning

(3), investigating long term consequences of adverse pregnancy outcomes is essential

in general, but more particular to check whether the same differential outcomes can be

observed (i.e., do the ethnicity related differential effects continue after birth). A remote

consequence of the research proposed under 1 to 3 could be different risk assessment

tools and strategies for different population strata.

• The financial remuneration structure should take into account the considerable regional

risk differences in the perinatal context; if regional budgeting is aimed at, adjusting for

maternal characteristics only is insufficient.

Recommendations Part II

• Improvement of hospital organisation deserves priority. Multiple factors appear

modifiable, e.g., 7*24h organisation including the timing of (planned) deliveries in a

fashion that problems most likely will emerge during daytime hours. Also, proactive

obstetric policy should be advocated, at least in the Dutch context. We expect policies

to already change towards a more proactive approach; however, no recent data are

available to substantiate this.

• Centralisation of acute obstetric care services should take local configuration and

measured hospital organisational features into account. Apart from the academic

perinatal centres, hospital size is not a decisive factor in this regard. Side effects of

centralisation may exceed the intended benefits.

203

GEN

ERAL D

ISCUSSIO

N

11

• Studies aimed at the prevention of preterm birth are expected to grossly affect perinatal

mortality. Rigorous anti-smoking policies combined with organisational and financial

support for active antenatal prevention are justified. Also, the ‘term’ threshold should

change from 37.0 to 38.0 weeks.

• In prioritising regions for public health interventions, crude and standardised outcome

both need to be made available.

• Home birth as an option should be restricted to low risk multiparous women, but is

only a sustained option if risk selection throughout pregnancy considerably improves.

General recommendations

• Risk factors should be detected in a consistent way (possibly through a standardised

checklist based approach) during antenatal risk selection preferably through a joint

effort by primary care community midwives and obstetricians.

• Mandatory information on lifestyle factors such as smoking should be incorporated in

detail in The Netherlands Perinatal Registry and the registration process.

• Data from medical and administrative registries should be conditionally made available

for research and care improving purposes, without prior individual consent, similar to,

e.g., the Nordic countries. Additionally, quality assurance should be a regular part, and

an intensive collaboration between registry authorities and hospitals and IT-stakeholders

for hospital administration systems is mandatory to respond to change.

204

GEN

ERAL D

ISCUSSIO

N

11

REFERENCES1. Timmermans S, Bonsel GJ, Steegers-Theunissen RP,

et al. Individual accumulation of heterogeneous risks explains perinatal inequalities within deprived neighbourhoods. Eur J Epidemiol. 2011; 26:165-80.

2. Burris HH, Collins JW, Jr. Race and preterm birth--the case for epigenetic inquiry. Ethn Dis. 2010 Summer;20:296-9.

3. Collins JW, Jr., David RJ, Rankin KM, Desireddi JR. Transgenerational effect of neighborhood pover-ty on low birth weight among African Americans in Cook County, Illinois. Am J Epidemiol. 2009 Mar 15;169(6):712-7.

4. Kramer MR, Hogue CR. What causes racial dispari-ties in very preterm birth? A biosocial perspective. Epidemiol Rev. 2009;31:84-98.

5. Fang J, Madhavan S, Bosworth W, Alderman MH. Residential segregation and mortality in New York City. Soc Sci Med. 1998 Aug;47:469-76.

6. Halpern D, Nazroo J. The ethnic density effect: results from a national community survey of England and Wales. Int J Soc Psychiatry. 2000 Spring;46:34-46.

7. Neeleman J, Wessely S. Ethnic minority suicide: a small area geographical study in south London. Psychol Med. 1999 Mar;29:429-36.

8. Rabkin JG. Ethnic density and psychiatric hos-pitalization: hazards of minority status. Am J Psychiatry. 1979 Dec;136:1562-6.

9. Grady SC. Racial disparities in low birthweight and the contribution of residential segregation: a multilevel analysis. Soc Sci Med. 2006 Dec;63: 3013-29.

10. Altman DG. Strategies for community health intervention: promises, paradoxes, pitfalls. Psy-chosom Med. 1995;57:226-33.

11. Buitendijk SE, Nijhuis JG. High perinatal mortality in the Netherlands compared to the rest of Europe. Ned Tijdschr Geneeskd. 2004;148:1855-60.

12. Mohangoo AD, Buitendijk SE, Hukkelhoven CW, et al. Higher perinatal mortality in The Netherlands than in other European countries: the Peristat-II study. Ned Tijdschr Geneeskd. 2008;152:2718-27.

13. Sheldon T. Perinatal mortality in Netherlands third worst in Europe. BMJ. 2008;337:a3118.

14. Fleischman AR, Oinuma M, Clark SL. Rethinking the definition of “term pregnancy”. Obstet Gy-necol. 2010;116:136-9.

15. Reddy UM, Bettegowda VR, Dias T, Yamada-Kushnir T, Ko CW, Willinger M. Term pregnancy: a period of heterogeneous risk for infant mortality. Obstet Gynecol. 2011;117:1279-87.

16. Sheldon T. Prompt access to expert care is among advice to reduce perinatal deaths in Netherlands. BMJ. 2010;340:c277.

17. Tomorrow’s specialist: working party report, September 2012. London: The Royal College of Obstetricians and Gynaecologists; 2012. (Online http://www.rcog.org.uk/files/rcog-corp/Tomor-row’s%20Specialist_FullReport.pdf ).

18. Heller G, Richardson DK, Schnell R, Misselwitz B, Kunzel W, Schmidt S. Are we regionalized enough? Early-neonatal deaths in low-risk births by the size of delivery units in Hesse, Germany 1990-1999. Int J Epidemiol. 2002;31:1061-8.

19. Hemminki E, Heino A, Gissler M. Should births be centralised in higher level hospitals? Experiences from regionalised health care in Finland. BJOG. 2011;118:1186-95.

20. Amelink-Verburg MP, Buitendijk SE. Pregnancy and labour in the Dutch maternity care system: what is normal? The role division between midwives and obstetricians. J Midwifery Womens Health. 2010;55:216-25.

21. Bais JMJ, Pel M. The basis of the Dutch obstetric system: risk selection. Eur Clinics Obstet Gynaecol. 2007;2:209-12.

22. Commisie Verloskunde CVZ. List Of Obstetric Indications (in Dutch: Verloskundig Vademecum) 2003. Diemen: 2003 (online http://www.knov.nl/docs/uploads/Verloskundig_Vademecum_2003.pdf).

23. Bonsel GJ, Birnie E, Denktas S, Poeran J, Steegers EAP. Dutch report: Lijnen in de Perinatale Sterfte, Signalementstudie Zwangerschap en Geboorte 2010. Rotterdam: Erasmus MC; 2010.

205

GEN

ERAL D

ISCUSSIO

N

11

24. Denktaş S, Waelput AJM, van Voorst S, et al. Healthy pregnancy 4 all: preconception care and risk selection during pregnancy (in Dutch: Precon-ceptiezorg en risicoselectie tijdens de zwanger-schap). Tijdschrift voor gezondheidswetenschap-pen. 2012;90:479-83.

25. Halliday J. WUDWAW ‘Who Usually Delivers Whom and Where’, Perinatal Data Collection Unit. Victoria, Australia: Victorian Government Department of Human Services; 1999.

26. Hampson JP, Roberts RI, Morgan DA. Shared care: a review of the literature. Fam Pract. 1996;13:264-79.

27. Tucker J, Florey CD, Howie P, McIlwaine G, Hall M. Is antenatal care apportioned according to obstetric risk? The Scottish antenatal care study. J Public Health Med. 1994;16:60-70.

28. Chan FY, Pun TC, Tse LY, Lai P, Ma HK. Shared antenatal care between family health services and hospital (consultant) services for low risk women. Asia Oceania J Obstet Gynaecol. 1993;19:291-8.

29. Chervenak FA, McCullough LB, Brent RL, Levene MI, Arabin B. Planned home birth: the professional responsibility response. Am J Obstet Gynecol. 2013;208:31-8.

30 . Evers AC, Brouwers HA, Hukkelhoven CW, et al. Perinatal mortality and severe morbidity in low and high risk term pregnancies in the Netherlands: prospective cohort study. BMJ. 2010;341:c5639.

31 . Evers AC, van Leeuwen J, Kwee A, et al. Mortality and morbidity among full-term neonates in a neonatal intensive care unit in the Utrecht region, the Netherlands. Ned Tijdschr Geneeskd. 2010;154:118.

32 . Gottvall K, Waldenstrom U, Tingstig C, Grunewald C. In-hospital birth center with the same medical guidelines as standard care: a comparative study of obstetric interventions and outcomes. Birth. 2011;38:120-8.

33 . de Neef T, Hukkelhoven CW, Franx A, van Ever-hardt E. Uit de lijn der verwachting. Nederl Tijdschrift Obstet Gynaecol 2009;122:34-342.

34 . Van der Velden J. Dutch report: Een goed begin. Veilige zorg rond zwangerschap en geboorte. Advies Stuurgroep Zwangerschap en Geboorte. The Hague: 2009.

35 . Moster D, Lie RT, Markestad T. Neonatal mortality rates in communities with small maternity units compared with those having larger maternity units. BJOG. 2001;108:904-9.

36 . Woo YL, Kyrgiou M, Bryant A, Everett T, Dickin-son HO. Centralisation of services for gynae-cological cancer. Cochrane Database Syst Rev. 2012;3:CD007945.

37 . Tracy SK, Sullivan E, Dahlen H, Black D, Wang YA, Tracy MB. Does size matter? A population-based study of birth in lower volume maternity hospitals for low risk women. BJOG. 2006;113:86-96.

38 . Phibbs CS, Bronstein JM, Buxton E, Phibbs RH. The effects of patient volume and level of care at the hospital of birth on neonatal mortality. JAMA. 1996;276:1054-9.

39 . Ravelli AC, Jager KJ, de Groot MH, et al. Travel time from home to hospital and adverse perinatal outcomes in women at term in the Netherlands. BJOG. 2011;118:457-65.

40 . Perinatal care in The Netherlands 2008 (in Dutch: Perinatale zorg in Nederland 2008). Utrecht: The Netherlands Perinatal Registry; 2011.

41 . Bonsel GJ, Poeran J, De Graaf JP, Borsboom GJJM, Birnie E, Steegers EAP, Mackenbach JP. Dutch report: Gevolgen van Concentratie van Perinatale Zorg. Rotterdam: Erasmus MC; 2012.

42 . Strand LB, Barnett AG, Tong S. The influence of season and ambient temperature on birth out-comes: a review of the epidemiological literature. Environ Res. 2011;111:451-62.

43 . Galan HL, Hussey MJ, Barbera A, et al. Relationship of fetal growth to duration of heat stress in an ovine model of placental insufficiency. Am J Obstet Gynecol. 1999;180(5):1278-82.

44 . Wells JC. Thermal environment and human birth weight. J Theor Biol. 2002;214(3):413-25.

45 . Reitsma JB. Registers in Cardiovascular Epide-miology. PhD thesis Academic Medical Center, University of Amsterdam ed; Amsterdam: 1999.

46 . Monson RR: Occupational Epidemiology. 2 edition. Boca Raton: CRC Press Inc; 1990, 72-74.

47 . Greenland S, Rothman KJ: Introduction to Cat-egorical Statistics in Modern Epidemiology. 2 edition. Philadelphia: Lippincott-Raven Publish-ers; 1998.

206

GEN

ERAL D

ISCUSSIO

N

11

48 . Bailit J, Garrett J. Comparison of risk-adjustment methodologies for cesarean delivery rates. Obstet Gynecol. 2003;102:45-51.

49 . Field A: Discovering Statistics Using SPSS. 3rd edition. London: Sage Publications; 2009.

50 . Twisk JWR. Applied Multilevel Analysis. 4 ed. Cambridge: Cambridge University Press; 2010.

51 . Figueras F, Gardosi J. Intrauterine growth restric-tion: new concepts in antenatal surveillance, di-agnosis, and management. Am J Obstet Gynecol. 2011 Apr;204:288-300.

52 . Commisie Verloskunde CVZ. List Of Obstetric Indications (in Dutch: Verloskundig Vademecum) 2003. Diemen: 2003 (online: http://www.knov.nl/docs/uploads/Verloskundig_Vademecum_2003.pdf).

12English and Dutch summary

208

12

ENG

LISH SU

MM

ARY

ENGLISH SUMMARYIn The Netherlands perinatal mortality rates exceed the European average. Also, within The

Netherlands inequalities in perinatal health are present, mainly in the four largest cities (‘G4’)

and deprived neighbourhoods. We state three levels of inequalities: (1) The Netherlands vs.

Europe, (2) regions and G4-cities vs. the remainder of The Netherlands, and (3) deprived vs.

non-deprived neighbourhoods and boroughs within these G4-cities. The causal factors for

perinatal inequalities may differ between the three geographic levels.

In response to the high perinatal mortality, the Dutch Minister for Health as well as the

municipality of Rotterdam issued several measures and research topics. This thesis reports

on local (PART I) and national (PART II) initiatives in the context of improving perinatal

health in The Netherlands. The aim is to investigate the main contributing factors in

adverse perinatal outcomes on the three geographic levels (international, national,

regional). Throughout this thesis, the concept of ‘Big4’ is applied. ‘Big4’ refers to four adverse

pregnancy outcomes (perinatal morbidities) which precede perinatal mortality in 85% of

cases: congenital anomalies, preterm birth (<37th week of gestation), small for gestational

age (SGA, birthweight below the 10th percentile for gestational age) or low Apgar score

(<7, 5 minutes after birth).

On the first geographic level, i.e., The Netherlands vs. Europe, the Dutch system of obstetric

care significantly differs from those in other Western countries, with a high percentage

of homebirths (around 20%), and an independent role for primary care community

midwives. Most important, however, is the risk-based level of care: primary care for low risk

pregnancies provided by independently practicing community midwives, and secondary/

tertiary care for high risk pregnancies provided by obstetricians in hospitals. Due to the

distinction between assumed low risk and high risk pregnancies, antenatal risk selection is

an essential indicator of the adequacy of the Dutch obstetric care system. We demonstrate

an insufficient separation of low risk pregnancies from a mix of low and high risk pregnancies

with a considerable part of this selection and subsequent referral taking place even during

parturition (chapter 7).

Home birth, as another unique feature, is generally not associated with increased intra-

partum and early neonatal mortality, under routine conditions (chapter 8). However, the

safety of home births is dependent on antenatal risk selection as only assumed low risk

women can be offered the choice of a birth at home. However, as antenatal risk selection

still results in 6% to 9% of high risk (‘Big4’) women in primary care, intended for low risk

pregnancies, improvement is mandatory (chapter 7).

209

12

ENG

LISH SU

MM

ARY

Chapters 5, 6, 9 and 10 pertain to perinatal health inequalities on the second geographic

level, i.e., regions and G4-cities vs. the remainder of the Netherlands. In chapter 5 we

demonstrate that an over 30% decrease in perinatal mortality is expected on the national

level through optimisation of (hospital) organisational features. This optimisation may be

achieved through centralisation of acute obstetric care services. However, hospitals appear

very heterogeneous in organisational features affecting perinatal outcome. Therefore, in

centralisation, these organisational features need to be taken into account as well as the

full scope of negatively (e.g., increased travel time) and positively (e.g., better 7*24h access

to specialised care) influencing factors (chapter 9). Due to the complexity of this matter

we agree with the statement from the Royal College of Obstetricians and Gynaecologists:

‘localised where possible, centralised where necessary’.

Chapter 6 reports on the selection of priority regions in which to implement intensified

preconception care and uniform antenatal risk selection. This selection was based on

indicators related to perinatal mortality, morbidity and healthcare factors. Both selecting

strategies (regarding to risk selection and preconception care) point out the four largest

cities (‘G4’, i.e., Amsterdam, Rotterdam, The Hague, Utrecht) as a priority region, thus

illustrating the increased risk for adverse perinatal outcome in large urban areas. G4-cities

surfacing in both selecting strategies emphasises the multifactorial effects on and the

complexity of urban perinatal health.

In chapter 10 we demonstrate the effect of climatological factors, i.e., seasonality, extremes

in temperature and cumulative sunshine exposure, on birthweight. Minimum tempera-

ture exposure in the second and third trimester was associated with higher birthweight,

maximum temperature exposure for all exposure windows was associated with lower

birthweight, and cumulative sunshine exposure was associated with higher birthweight.

On the population level, significant regional differences in birthweight exist, most likely

attributable to particularly maximum temperatures, with moderation of the effect in coastal

areas. For healthy babies born at term these differences in birthweight appear small; how-

ever, more severe detrimental effects may emerge for vulnerable subgroups, e.g., women

carrying growth restricted babies.

Chapters 2, 3 and 4 pertain to perinatal health inequalities on the third geographic level,

i.e., deprived vs. non-deprived neighbourhoods and boroughs within G4-cities, in particular

Rotterdam. We demonstrate large differences in absolute perinatal mortality and perinatal

morbidity rates between neighbourhoods within the city of Rotterdam with perinatal

mortality rates as high as 37 per 1,000 births (chapter 2). These inequalities remain after

210

12

ENG

LISH SU

MM

ARY

standardisation, also implying differences in possible causes per borough with subsequent

different policy measures per borough (chapter 3). Common mechanisms of adverse

perinatal outcome in deprived neighbourhoods are thought to be social deprivation and

accumulation of risks for perinatal mortality and morbidity. In chapter 4 we demonstrate

differential effects of social deprivation on Western and non-Western women in Rotterdam.

Improvement in neighbourhood social quality causes improvement in perinatal outcomes

for Western women only. Policy measures aimed at improving social quality for Western

women may improve outcome in Western women, however, alternative approaches may

be necessary for non-Western groups.

The most important recommendations for improvement of perinatal health on the first

geographic level include improvement of risk selection (possibly through a standardised

checklist based approach) and ‘shared care’ as the success of improvement strategies

depends on a joint collaboration of midwives and obstetricians. Important recommendations

on the second geographic level of perinatal health inequalities include a sensible policy

on centralisation of acute obstetric care and additional research and more public health

awareness for the complex issue of urban perinatal health. On the third geographic

level of perinatal health, we propose additional research on the differential effects of

social deprivation on Western and non-Western women. Our results imply strategies to

improve social quality for in particular for Western women and more focus on alternative

approaches for non-Western groups; these alternative approaches may be based on results

from additional research. Furthermore, neighbourhood, and borough specific policies are

mandatory on tackling urban perinatal health inequalities.

211

12

NED

ERLAN

DSE SA

MEN

VATTING

NEDERLANDSE SAMENVATTINGNederland kent een van de hoogste perinatale sterftecijfers in Europa met ook binnen

Nederland ongelijkheid in perinatale gezondheid waarbij de meest ongunstige uitkomsten

gezien worden in de vier grootste steden (G4: Amsterdam, Rotterdam, Den Haag, Utrecht) en

achterstandswijken. De ongelijkheid in perinatale gezondheid kan onderverdeeld worden

naar drie niveaus gebaseerd op geografie: (1) Nederland tegenover Europa, (2) de G4 en

regio’s binnen Nederland, en (3) achterstandswijken tegenover niet-achterstandswijken. De

oorzaken voor de ongelijkheid in perinatale gezondheid verschillen hoogstwaarschijnlijk

per niveau.

De hoge perinatale sterfte in Nederland werd zowel voor de landelijke als regionale poli-

tiek een prioriteit met voorstellen van de minister van Volksgezondheid en de gemeente

Rotterdam tot onderzoek en beleidsmaatregelen. Dit proefschrift rapporteert over regio-

nale (DEEL I) en landelijke initiatieven (DEEL II) ter verbetering van de perinatale sterfte in

Nederland. Hoofddoel van het proefschrift is dan ook de belangrijkste factoren te iden-

tificieren die bijdragen aan de slechte perinatale gezondheid in Nederland betreffende

de drie geografische niveaus (internationaal, nationaal, regionaal). Het concept van ‘Big4’

aandoeningen speelt een belangrijke rol in de stukken behorend bij dit proefschrift. ‘Big4’

staat voor de vier ongunstige zwangerschapsuitkomsten die voorafgaan aan 85% van alle

gevallen van perinatale sterfte: aangeboren afwijkingen, vroeggeboorte (vóór 37 weken

zwangerschap), te laag geboortegewicht voor de zwangerschapsduur (geboortegewicht

<p10), en een lage Apgarscore (<7, 5 minuten na de geboorte).

Voor de vergelijking op het eerste geografische niveau (Nederland tegenover de rest van

Europa) speelt het unieke systeem van verloskundige zorg in Nederland een belangrijke

rol; uniek in die zin dat, in vergelijking met andere Westerse landen, in Nederland een hoog

percentage van de zwangere vrouwen thuis bevalt (rond de 20%). Een ander verschil met

andere Westerse landen is de belangrijke rol van de eerstelijns verloskundige. Het belang-

rijkste verschil is echter het intrinsieke verschil tussen hoogrisico- en laagrisicozwanger-

schappen met daarbij behorende risicogeleide zorg in de verloskunde: eerstelijns zorg voor

laagrisicozwangerschappen verleend door verloskundigen, en tweede-/derdelijns zorg

voor hoogrisicozwangerschappen in ziekenhuizen verleend door gynaecologen. Door dit

onderscheid en de specifieke bijbehorende zorgconsequenties speelt antenatale risicose-

lectie een belangrijke rol in het Nederlandse systeem en kan dan ook gezien worden als

een belangrijke kwaliteitsindicator: ‘hoe goed worden laagrisico- en hoogrisicozwanger-

schappen van elkaar onderscheiden’. In dit proefchrift tonen wij aan dat deze twee groepen

212

12

NED

ERLAN

DSE SA

MEN

VATTING

suboptimaal kunnen worden onderscheiden van elkaar in de periode vóór de bevalling.

Daarnaast wordt ook een substantieel deel van de zwangere vrouwen gedurende de baring

aangeduid zijnde hoogrisico en daarop doorverwezen naar het ziekenhuis (hoofdstuk 7).

De thuisbevalling is een ander uniek kenmerk van het Nederlandse verloskundige systeem.

Onder normale omstandigheiden is deze niet geassocieerd met een verhoogde sterftekans

tijdens de baring of in de eerste week na de bevalling (hoofdstuk 8). Echter, de veiligheid

van thuisbevallingen hangt sterk af van herkenning en selectie van risico’s vóór de start

van de baring omdat alleen vrouwen die geacht worden laagrisico te zijn de keus wordt

geboden om thuis te bevallen. Door suboptimale risicoselectie bevalt nog steeds een niet

onbelangrijk deel van de hoogrisico (‘Big4’) vrouwen in de eerste lijn: 6% tot 9%. Het is van

cruciaal belang dit percentage omlaag te krijgen aangezien eerstelijns zorg uitsluitend

toegerust is op een laagrisicozwangerschap en -bevalling (hoofdstuk 7).

Hoofdstukken 5, 6, 9 en 10 behandelen verschillen in perinatale gezonheid op het tweede

geografische niveau: de G4 en regio’s binnen Nederland. In hoofdstuk 5 tonen wij aan dat

door optimalisatie van ziekenhuisorganisatie de perinatale sterfte met mogelijk meer dan

30% zou kunnen dalen. Een veelgenoemde beleidsmaatregel in dit kader is centralisatie

van acute verloskundige zorg, wat geacht wordt organisatieaspecten van ziekenhuizen

te verbeteren door het aantal ziekenhuizen dat acute verloskundige zorg biedt sterk te

verminderen. Dit impliceert sluiting van kleinere ziekenhuizen waarbij er een geringer

aantal grotere ziekenhuizen overblijft. Er bestaan echter grote verschillen tussen zieken-

huizen voor wat betreft organisatieaspecten met op hun beurt een positief of negatief

effect op de perinatale sterfte. Dit is een belangrijke factor waar rekening mee gehouden

moet worden bij een beleid van centralisatie van acute verloskundige zorg. Een verstandig

centralisatiebeleid houdt verder rekening met het hele spectrum van factoren die zowel

een positief (bijvoorbeeld 7*24 toegang tot specialistische zorg) als negatief (bijvoorbeeld

een toegenomen reistijd naar het ziekenhuis) effect kunnen hebben op de perinatale ge-

zondheid (hoofdstuk 9). Het centralisatievraagstuk is erg complex; wij ondersteunen dan

ook het standpunt van de Britse Royal College of Obstetricians and Gynaecologists: ‘localised

where possible, centralised where necessary’.

In hoofdstuk 6 wordt ingegaan op twee selectiestrategieën ter identificatie van hoogrisico-

gebieden in Nederland met als doel hier intensieve preconceptiezorg of uniforme antena-

tale risicoselectie toe te passen. Deze selectie is gebaseerd op indicatoren gerelateerd aan

perinatale sterfte, morbiditeit en zorgfactoren. De selectiestrategieën identificeerden de G4

als hoogrisico voor zowel preconceptiezorg als antenatale risicoselectie; hierbij werd nog

213

12

NED

ERLAN

DSE SA

MEN

VATTING

eens bevestigd dat er een verhoogd risico bestaat op ongunstige zwangerschapsuitkom-

sten in de G4. Daarnaast was dit ook een bevestiging van de complexiteit en multifactoriële

oorzaken van perinatale gezondheid in de grote steden.

In hoofdstuk 10 tonen wij effecten van het klimaat, te weten seizoen, temperatuurextremen

en cumulatieve blootstelling aan zonlicht, op geboortegewicht. Blootstelling aan minimum

temperaturen in het tweede en derde trimester van de zwangerschap was geassocieerd

met een hoger geboortegewicht; blootstelling aan maximum temperaturen gedurende de

periconceptionele periode, alle trimesters van de zwangerschap en op de dag van geboorte

was geassocieerd met een lager geboortegewicht; cumulatieve blootstelling aan zonlicht

gedurende alle trimesters was geassocieerd met een hoger geboortegewicht. Ook bleken

er binnen Nederland substantiële klimaatgerelateerde regionale verschillen te bestaan in

geboortegewicht die waarschijnlijk voor het belangrijkste deel toegeschreven kunnen wor-

den aan de blootstelling aan maximum temperaturen. Voor gezonde á terme zuigelingen

lijken deze verschillen klein; echter, de effecten kunnen ernstiger zijn voor risicogroepen

zoals vrouwen die zwanger zijn van groeivertraagde baby’s.

Hoofdstukken 2, 3 en 4 omvatten verschillen in perinatale gezondheid op het derde

geografische niveau: achterstandswijken tegenover niet-achterstandswijken binnen de

G4, in het bijzonder Rotterdam. Er bestaan grote verschillen in absolute perinatale sterfte

en perinatale morbiditeit tussen verschillende wijken binnen de stad Rotterdam met peri-

natale sterftecijfers oplopend tot 37 per 1.000 geboorten (hoofdstuk 2). Dat deze verschil-

len ook na standaardisatie aanhouden impliceert dat er verschillende oorzaken bestaan

per deelgemeente waarbij deelgemeente-specifiek beleid is geïndiceerd (hoofdstuk 3).

Veelgenoemde mechanismen van ongunstige zwangerschapsuitkomsten in achterstands-

wijken zijn sociale achterstand en cumulatie van risico’s gerelateerd aan perinatale sterfte

en morbiditeit. In hoofdstuk 4 tonen wij differentiële effecten van sociale achterstand op

Westerse en niet-Westerse zwangere vrouwen in Rotterdam. Met verbetering van sociale

kwaliteit in de buurt verbetert ook de perinatale gezondheid van Westerse vrouwen; een

verband dat niet gezien werd voor niet-Westerse vrouwen. Beleidsmaatregelen gericht op

verbetering van sociale kwaliteit, in het bijzonder gericht op Westerse vrouwen in achter-

standssituaties, zal waarschijnlijk ook de perinatale gezonheid in deze groep ten goede

komen. Voor niet-Westerse vrouwen is waarschijnlijk een alternatieve aanpak nodig.

Verbetering van perinatale gezondheid op het eerste geografische niveau zal waarschijn-

lijk vooral bewerkstelligd kunnen worden door verbetering van antenatale risicoselectie

(mogelijk door een checklist-gebaseerde aanpak) en zogenaamde shared care waarbij

214

12

NED

ERLAN

DSE SA

MEN

VATTING

risicoselectie een gemeenschappelijke verantwoordelijkheid is van zowel verloskundigen

als gynaecologen. Belangrijke aanbevelingen op het tweede geografische niveau omvatten

een verstandig beleid van centralisatie van acute verloskundige zorg, en nader onderzoek

naar het complexe thema van grootstedelijke perinatale gezondheid met meer focus op

de maatschappelijke gezondheidszorg. Voor wat betreft het derde geografische niveau is

de belangrijkste aanbeveling het nader onderzoek naar differentiële effecten van sociale

achterstand op Westerse en niet-Westerse vrouwen. Onze studieresultaten impliceren een

beleid ter verbetering van sociale kwaliteit voor met name Westerse vrouwen met een focus

op alternatieve maatregelen voor niet-Westerse vrouwen. Deze alternatieve maatregelen

zullen moeten volgen uit resultaten van nader onderzoek. Een andere belangrijke aanbeve-

ling op dit niveau is buurt- en deelgemeente-specifiek beleid voor wat betreft verbetering

van grootstedelijke perinatale gezondheid.

13Authors and affiliations

Manuscripts About the author

PhD portfolioWord of thanks

216

13AU

THO

RS AN

D A

FFILIATION

S

AUTHORS AND AFFILIATIONSFrom the Department of Obstetrics and Gynaecology, Division of Obstetrics and Prenatal

Medicine, Erasmus MC, University Medical Centre Rotterdam, The Netherlands

Semiha Denktaş, Eric A.P. Steegers, Gouke J. Bonsel, Arno F.G. Maas, Johanna (Hanneke)

P. de Graaf, Adja J.M. Waelput, Lieke C. de Jong-Potjer, Sabine F. van Voorst, Amber A. Vos,

Jacoba van der Kooy

From the Department of Public Health, Erasmus MC, University Medical Centre Rotterdam,

The Netherlands

Gouke J. Bonsel, Gerard J.J.M. Borsboom, Johan P. Mackenbach

From the Midwifery Academy, Rotterdam University of Applied Sciences, Rotterdam, The

Netherlands

Gouke J. Bonsel

From the Institute of Health Policy and Management, Erasmus University, Rotterdam, The

Netherlands

Erwin Birnie

From the Municipal Public Health Authority ‘GGD Rotterdam Rijnmond’, Rotterdam, The

Netherlands

H.J. (Ernie) van der Weg, A.J.J. (Toon) Voorham

217

13

MA

NU

SCRIPTS

MANUSCRIPTS

Manuscripts related to this thesis

Chapter 2

J. Poeran, S. Denktaş, E. Birnie, G.J. Bonsel, E.A.P. Steegers. Urban perinatal health inequalities.

Journal of Maternal-Fetal and Neonatal Medicine 2011;24:643-6.

Chapter 3

J. Poeran, S. Denktaş, E. Birnie, H.J. van der Weg, A.J.J. Voorham, E.A.P. Steegers, G.J.

Bonsel. Pregnancy and birth in Rotterdam: a local health report. Tijdschrift voor

gezondheidswetenschappen. 2012;90:496-503. (Article in Dutch, translated into English

for this thesis.)

Chapter 4

J. Poeran, A.F.G. Maas, E. Birnie, S. Denktaş, E.A.P. Steegers, G.J. Bonsel. Social deprivation

and adverse perinatal outcomes among Western and non-Western pregnant women in a

Dutch urban population. Social Science and Medicine;2013;83:42-49.

Chapter 5

J. Poeran, G.J.J.M. Borsboom, J.P. de Graaf, E. Birnie, E.A.P. Steegers, G.J. Bonsel. Population

attributable risks of patient, child and organisational risk factors for perinatal mortality.

Manuscript submitted for publication.

Chapter 6

S. Denktaş, J. Poeran, A.J.M. Waelput, L.C. de Jong-Potjer, S.F. van Voorst, A.A. Vos, G.J. Bonsel,

E.A. P. Steegers. The ‘Healthy Pregnancy 4 All’ Study: Design and cohort profile. Manuscript

submitted for publication.

Chapter 7

J. Poeran, J. van der Kooy, E. Birnie, S. Denktaş, E.A.P. Steegers, G.J. Bonsel. The effectiveness

of risk selection in the Dutch obstetric care system. Manuscript submitted for publication.

218

13M

AN

USCRIPTS

Chapter 8

J. van der Kooy, J. Poeran, J.P. de Graaf, E. Birnie, S. Denktaş, E.A.P. Steegers, G.J. Bonsel.

Planned home compared with planned hospital births in The Netherlands: intrapartum and

early neonatal death in low-risk pregnancies. Obstetrics and Gynecolology 2011;118:1037-

46.

Chapter 9

J. Poeran, G.J.J.M. Borsboom, J.P. de Graaf, E. Birnie, E.A.P.Steegers, J.P. Mackenbach, G.J.

Bonsel. Does centralisation of acute obstetric care reduce perinatal mortality? An empirical

study on pros and cons in The Netherlands. Manuscript submitted for publication.

Chapter 10

J. Poeran, E. Birnie, E.A.P. Steegers, G.J. Bonsel. The impact of extremes in outdoor

temperature and sunshine exposure on birthweight. Manuscript submitted for publication.

Other manuscripts

Scientifi c journals

S.J.A.L. Hemelaar, J. Poeran, E.A.P. Steegers, W.I. van der Meijden. Neonatal herpes infections

in The Netherlands in the period 2006-2011. Manuscript submitted for publication.

J. de Graaf, J. Schutte, J. Poeran, J. Van Roosmalen, G.J. Bonsel, E.A.P. Steegers. Regional

differences in Dutch maternal mortality. BJOG. 2012;119:582-8.

J.P. de Graaf, J.M. Schutte, J. Poeran, J. Van Roosmalen, G.J. Bonsel, E.A. Steegers. Regional

differences in maternal mortality in the Netherlands. Ned Tijdschr Geneeskd. 156:A4952.

(in Dutch with English abstract)

J. Poeran, E.A.P. Steegers, G.J. Bonsel. Approach to perinatal mortality in the Netherlands:

outcomes of a systematic expert study. Ned Tijdschr Geneeskd. 2012;155:A4499. (in Dutch

with English abstract)

J. Poeran, H. Wildschut, M. Gaytant, J. Galama, E.A.P. Steegers, W. van der Meijden. The

incidence of neonatal herpes in The Netherlands. J Clin Virol. 2008;42:321‐5.

219

13

MA

NU

SCRIPTS

Health reports

G.J. Bonsel, J. Poeran, G.J.J.M. Borsboom, J.P. de Graaf, E. Birnie, E.A.P. Steegers. ‘Populatie

attributieve risico’s van moeder-, kind- en organisatie factoren voor perinatale sterfte’ The

population attributable risk of mother-, child- and organisational factors for the perinatal

mortality in The Netherlands 2000‐2008. Rotterdam: Erasmus MC, 2012. (Report for ‘ZonMw’

The Netherlands Organisation for Health Research and Development, report in Dutch with

English abstract)

G.J. Bonsel, J. Poeran, J.P. de Graaf, G.J.J.M. Borsboom, E. Birnie, E.A.P. Steegers, J.P. Macken-

bach. ‘De effecten van concentratie van verloskundige zorg’ Merging hospitals with acute

obstetric care in The Netherlands: effects on perinatal mortality in a system with high

intrapartum referral rates and heterogeneous hospitals. Rotterdam: Erasmus MC, 2012.

(Report for a group of 40 small and medium sized hospitals in The Netherlands which was

presented to the Dutch Minister for Health, report in Dutch with English abstract)

J. Poeran, E. Birnie, S. Denktaş, E.A.P. Steegers, G.J. Bonsel. ‘Zwangerschap en geboorte in

Rotterdam: deelgemeenterapportage’ Perinatal health in Rotterdam 2000-2007. Rotterdam:

Erasmus MC, 2011. (Report for the Rotterdam city Council, report in Dutch with English

abstract)

G.J. Bonsel, E. Birnie, S. Denktaş, J. Poeran, E.A.P. Steegers. ‘Signalementstudie: lijnen in de

perinatale sterfte’ National report on priority topics in perinatal health improvement in The

Netherlands. Rotterdam: Erasmus MC, 2010. (Report for ‘ZonMw’ The Netherlands Organi-

sation for Health Research and Development, presented to the Dutch Minister for Health,

report in Dutch with English abstract)

220

13A

BOU

T THE AU

THO

R

ABOUT THE AUTHORAs the mother of the author, I am very honoured in helping him write this piece for his PhD-

thesis. Virender Jashvant John Poeran was born at a gestational age of 38 weeks on March

21st, 1983 in the Diakonessen hospital in Paramaribo, Suriname. Labour progressed fast, and

early in the morning at exactly 06:40 AM he came into this world in vertex presentation with

a birthweight of 3,560 grams, over 500 grams heavier than his older brother. His height was

49 centimetres with a head circumference of 36.5 centimetres. Right after birth, this already

attention seeking baby started crying loudly after which the midwife gave him an Apgar

score of 10/10. During my whole pregnancy I had expected a girl and I even came up with

the name ‘Sheila’. Having the nurse yelling ‘it’s a boy!’ made me adjust my reality a bit, but I

soon fell in love with that beautiful smile. The most important thing I remember about his

childhood is that he was the naughtiest of my three children. I could not leave him alone for

one second. He was always up to doing something and always busy bullying his little sister.

Jashvant’s first career choices involved being a magician, a pilot and an astronaut. It wasn’t

until his late teens he started thinking of going to medical school for which he was accepted

in the summer of 2001. During his medical training he stumbled upon the path of research

in which he was guided by professor Steegers from the department of Obstetrics and dr.

van der Meijden from the department of Dermatology. At the end of medical school he

more or less made the decision to pursue a career in Obstetrics and Gynaecology. He

finished two clinical internships abroad (Thailand and Suriname) and one in the Albert

Schweitzer hospital in Dordrecht. After obtaining his medical degree in 2008 he worked

a short period as a physician at the department of Obstetrics and Gynaecology at the

Albert Schweitzer hospital in Dordrecht.

Six months later Jashvant was accepted for a PhD-program in Perinatal Epidemiology

under guidance of professors Bonsel and Steegers at Erasmus MC. This municipal

program in the city of Rotterdam had been initiated to reduce the high perinatal mortality

in the city. He was additionally granted the possibility to obtain a Master’s degree in

Epidemiology, which he finished in August 2012. It was during his PhD-period that Jashvant

was extensively re-thinking his career choices and decided to pursue a career in Psychiatry

instead of Obstetrics and Gynaecology. At the end of his PhD-research he therefore worked

as an emergency physician in Psychiatry at the Parnassia Bavo Group, a large suburban

mental hospital in the Rotterdam area.

Always the adventurer and a big traveller, at the end of his PhD-period Jashvant fell in

love with the city of New York. After several meetings with his future boss and mentor dr.

221

13

ABO

UT TH

E AUTH

OR

Mazumdar he decided to pursue his career in the United States instead of The Netherlands.

In April 2013 he started working as a postdoctoral associate of Epidemiology at

Weill Cornell Medical College in New York under guidance of dr. Mazumdar and dr.

Memtsoudis. He is currently studying for his medical degree to be recognised in the

United States and may someday enter clinical medicine again. Being in New York, he has

also signed up to run the New York City Marathon on November 3, 2013, for which he

will raise money for charity. Although I sometimes cannot follow his choices in life I try to

support him wherever I can; in the end, I truly can say I am a proud mother.

Chandra Kirpal, Jashvant’s mother

222

13PH

D PO

RTFOLIO

PHD PORTFOLIO

Summary of PhD training and teaching activities

Name PhD candidate:

Erasmus MC Department:

Research school:

PhD period:

Promotors:

Copromotors:

Virender Jashvant John PoeranObstetrics & GynaecologyNetherlands Institute for Health Sciences (NIHES)August 2009 - December 2012Prof. dr. Gouke J. BonselProf. dr. Eric A.P. SteegersDr. Erwin BirnieDr. Semiha Denktaş

Year Workload

(ECTS)

1. PHD TRAINING

General and specific courses

Master of Science in Epidemiology, Netherlands Institute for Health Sciences, Rotterdam, The Netherlands (NIHES)

2010-2012 35

Introduction in ‘R’ programming 2012 0.1Introduction in SPSS Data Entry 2010 0.3

International conferences

Symposium ‘How Maternity Care Develops in a Multi-Ethnic Urban Society’, Rotterdam, The Netherlands: oral presentation Urban perinatal health in Rotterdam

2012 1.5

59th Annual Meeting of the Society for Gynecologic Investigation, San Diego, USA: poster presentation Social deprivation and adverse perinatal outcomes among Western and non-Western pregnant women in a Dutch urban population

2012 1

58th Annual Meeting of the Society for Gynecologic Investigation, Miami, USA: poster

presentation The effectiveness of risk selection in the Dutch obstetric care system2011 1

11th Congress of the European Society of Contraception and Reproductive Health, The Hague, The Netherlands: oral presentation Unexpected large perinatal health inequalities in an urban, multi‐ethnic population in the Netherlands

2010 1.5

57th Annual Meeting of the Society for Gynecologic Investigation, Orlando, USA: poster presentation Unexpected high levels of perinatal health inequalities in an urban multi‐ethnic population in The Netherlands

2010 1

National conferences

Symposium ‘Urban Perinatal Health’ (in Dutch: ‘Grootstedelijke Perinatale Gezondheid’), Rotterdam, The Netherlands: oral presentation The Social Index and Urban Perinatal Health in Rotterdam (Assisted in organising this symposium)

2011 3

NCVGZ Dutch Annual Public Health Conference (In Dutch: ‘Nationaal Congres Volksgezondheid’), Amsterdam, The Netherlands: oral presentation The Social index and perinatal outcomes in the city of Rotterdam

2011 2

NCVGZ Dutch Annual Public Health Conference (In Dutch: ‘Nationaal Congres Volksgezondheid’), Rotterdam, The Netherlands: oral presentation Large perinatal health inequalities within the city of Rotterdam

2010 2

223

13

PHD

PORTFO

LIO

Year Workload

(ECTS)

Seminars, workshops, research meetings

The Netherlands Perinatal Registry Annual Meetings (‘PRN Contactdag’), Utrecht, The Netherlands

2009-2012 0.8

Attending weekly and quarterly research meetings of the Department of Obstetrics and Gynaecology (and Urology) with three oral presentations, Rotterdam, The Netherlands

2009-2012 5

Attending the annual meetings of the Collaboration of Rotterdam Regional Gynaecologists’ Teaching Hospitals (in Dutch: RGOC) ‘Wladimiroff Symposium with one oral presentation, Rotterdam, The Netherlands

2010-2012 2

Attending the symposium on behalf of the 2.5 year anniversary of the ‘Sophia Birth Centre’, Rotterdam, The Netherlands

2012 0.2

Attending the symposium ‘Home Birth or Hospital Birth’, Maastricht, The Netherlands 2012 0.3Visit from professor Eileen Hutton and staff, VU Medical Centre, Amsterdam, The Netherlands; oral presentation

2012 0.5

Visit from professor Lucilla Poston and staff, King’s College, London, United Kingdom; oral presentation

2011 0.5

Regional Obstetric Collaboration meeting (‘VSV’), Schiedam, The Netherlands with one oral presentation

2010 1

Other: Grant proposal assistance

Assisted in the writing of several grant proposals for ‘ZonMw’ The Netherlands Organisation for Health Research and Development

2009-2012 5

Other: writing of municipal and national (public) health reports

(Assisted in) writing of several national and municipal health reports for local and national (health) authorities: (1) ‘ZonMw’, (2) a collaboration of 40 small hospitals, (3) the Ministry of Health, (4) the Municipality of Rotterdam

2009-2012 15

2. TEACHING ACTIVITIES

Lecturing

Lectures and practicals on ‘Perinatal Health’ in The Netherlands and Rotterdam for Medical students; minor ‘Circle of Life’

2011-2012 5

Lectures on ‘Perinatal Health and Policy Implications’ for Clinical Midwives aspiring to obtain a Master’s degree as Physician Assistant

2011-2012 2

Lecture on ‘Ethnicity and Perinatal Health’ for NIHES Master’s candidates; course ‘Urban Perinatal Health and Health Care’

2012 1

Lecture for students doing their research electives from the Rotterdam University of Applied Sciences on ‘The Dutch System of Obstetric Healthcare’

2011 1

Tutoring

Tutoring first-year Medical students in the first four months of their studies, also preparing them for an academic environment

2012 1

Supervising (Master’s) theses

Medical student Arno F.G. Maas, title: ‘The Social Index and Adverse Perinatal Outcomes in Rotterdam’

2010 2

Medical student Elva Hinborch, title: ‘Environmental Factors Influencing Perinatal Health’ 2010 1Medical student Rénuka S. Birbal, title: ‘Congenital Anomalies in the Borough of Hoek van Holland’

2011 2

Medical student Steffannie J.A.L. Hemelaar, title: ‘The Incidence of Neonatal Herpes in The Netherlands 2005-2012’

2012 2

224

13W

ORD

OF TH

AN

KS

WORD OF THANKSDat ik hier vandaag sta, de laatste dag als promovendus, heb ik aan meerdere personen

te danken. Ik heb mijn promotie-onderzoek ervaren als een tijd waarin ik een enorme

persoonlijke en professionele groei heb doorgemaakt. Wetenschap bleek mij meer te

trekken dan ik aanvankelijk had durven denken, en ik kreeg er nog betaald voor ook.

Ondanks dat ik mijn toekomst al langere tijd zag als gynaecoloog, heb ik meerdere moeilijke

keuzes durven nemen met een enorme impact op mijn toekomstbeeld, maar die mij

uiteindelijk wel gelukkiger hebben gemaakt.

Allereerst wil ik iedereen bedanken die in een bepaalde vorm heeft bijgedragen aan de

totstandkoming van dit proefschrift, in het bijzonder mijn promotoren prof. dr. Gouke

Bonsel en prof. dr. Eric Steegers.

Gouke, ik herinner me onze eerste ontmoeting nog heel goed, je had heel wat te vertellen

over ene ‘Signalementstudie’, omdat ik net uit een nachtdienst kwam is het meeste langs me

heen gegaan. Dat was gelukkig niet het geval toen ik ook daadwerkelijk onder jouw hoede

op de afdeling kwam. Het is niet te meten hoeveel ik van je heb geleerd over epidemiologie,

statistiek, algemene wetenschap, en de meest actuele (maar ook de minder actuele)

roddels. De manier waarop je altijd het onderste uit de kan wil halen, altijd nieuwsgierig

bent naar hoe dingen in elkaar zitten, heeft erg inspirerend gewerkt. Ik ben erg blij dat jij

mijn promotor bent geweest, dank voor de inspirerende jaren.

Beste Eric, na mijn onderzoek als student geneeskunde bleef ik je per e-mail stalken omdat

ik graag gynaecoloog wilde worden en wilde promoveren op jouw afdeling. Dat laatste heb

ik inmiddels bereikt en voor wat betreft dat eerste heb ik een van de moeilijkste gesprekken

uit mijn leven met jou gevoerd. Dank voor de voortdurende steun, het begrip, maar vooral

ook het getoonde vertrouwen. Ik hoop dat we in de toekomst nog zullen samenwerken.

Mijn eerste copromotor dr. Erwin Birnie. Beste Erwin, een man van initieel weinig woorden,

maar wel de juiste. Jouw altijd scherpe commentaren op mijn manuscripten waren altijd

een motivator om mezelf te blijven verbeteren en altijd kritisch te blijven; dit waardeer

ik enorm. Iets waar ik altijd aan zal terugdenken zijn de ‘soep-momenten’ en later de ‘vis-

momenten’ waar altijd weer de nieuwste roddels werden uitgewisseld.

Mijn tweede copromotor dr. Semiha Denktaş. Beste Semiha, ook met jou is het altijd

prettig samenwerken geweest; met name de informele sfeer heb ik erg gewaardeerd.

Ik kon met alles naar je toe komen, je bood altijd een luisterend oor en je gaf ook altijd

advies hoe te handelen in moeilijke situaties. Ik heb erg veel respect voor hoe jij bent

225

13

WO

RD O

F THA

NKS

gepromoveerd, een groot gemeentelijk programma hebt geleid, een groot landelijk

programma leidt en daarnaast nog eens een moeder van twee kinderen bent. Ook aan

jouw doorzettingsvermogen en multifunctionaliteit hoop ik me eens te meten.

Dit hele promotietraject was niet mogelijk geweest zonder financiële ondersteuning van

de Gemeente Rotterdam in de vorm van het initialiseren van het programma ‘Klaar voor

een Kind’. Rotterdam heeft met dit project nog maar eens bewezen vooral te ‘doen’ en te

‘durven’ in plaats van te ‘dralen’. In dit kader ben ik ook de GGD Rotterdam Rijnmond erg

erkentelijk voor de samenwerking binnen dit programma, in het bijzonder drs. H.J. (Ernie)

van der Weg en dr. A.J.J. (Toon) Voorham.

Bij de realisatie van dit proefschrift is voornamelijk gebruik gemaakt van gegevens uit de

Perinatale Registratie Nederland (PRN). Graag wil ik dan ook bedanken het PRN bestuur,

drs. G. de Winter en natuurlijk dr. ir. C.W.P.M. (Chantal) Hukkelhoven.

Graag wil ik ook bedanken de overige leden van de leescommissie voor hun kritische blik

op mijn proefschrift: prof. dr. ir. A. Burdorf, prof. dr. A.J. van der Heijden en prof. dr.

A.P. Verhoeff.

Ook de leden van de grote commissie wil ik hierbij bedanken voor hun tijd, hun kritische

blik en de discussie over de inhoud van mijn proefschrift.

De ‘Klaar voor een Kind(eren)’ en de ‘Healthy Pregnancies’: Sevilay Temel, Chantal Quispel,

Ingrid Peters, Daniëlle van Veen, Mieke van Veen, Hanneke Torij, Kirsten Heetkamp, Isabel

Fino Dos Santos Boialvo, Mijke van den Berg, Tom Schneider, Anke Posthumus, Amber Vos,

Sabine van Voorst, Vera Schölmerich, Adja Waelput, Lieke de Jong-Potjer, prof. Merkus en de

rest van de twee teams die ik misschien nog vergeten ben. Dank voor de fijne samenwerking,

maar vooral ook de gezelligheid!

In het bijzonder wil ook bedanken Jolanda Claessens en Brenda Karsan die me altijd

wisten te helpen als ik even niet meer wist hoe ik iets geregeld moest krijgen.

Ook ben ik dr. J.P. (Hanneke) de Graaf erg erkentelijk voor het altijd brengen van een (niet

bescheiden) lach. Daarnaast heb ik een erg fijne tijd gehad in die hele mooie kamer in het

Hs-gebouw die je toch maar weer even voor ons geregeld had.

Drs. G.J.J.M. (Gerard) Borsboom; ik zal niet gauw die beruchte vrijdagavond vergeten

toen we kort voor een deadline een analysefout ontdekten. Ik weet zeker dat ik er een paar

grijze haren aan heb overgehouden. Ik heb mijn tijd met jou als erg prettig ervaren, je hebt

me leren programmeren in SAS en je hebt was nooit te beroerd om me te helpen als ik

weer eens wat statistiekvraagjes had. Dank voor de gezellige, en vooral ook leerzame tijd.

226

13W

ORD

OF TH

AN

KS

De eerste helft van ‘Babsvant’, Babs, mijn steun en toeverlaat zowel tijdens werk maar ook

ver daarbuiten. Wie had gedacht dat we het zo goed met elkaar zouden vinden, ik niet

in ieder geval. Je hebt me kritisch leren kijken naar mijzelf, we hebben veel filosfofische

gesprekken gevoerd over het geloof en van alles en nog wat. Ik weet nog steeds niet precies

wat ons nou zo bindt maar ik hoop dat we nog een heleboel in de toekomst samen zullen

doen.

Mijn vrienden, mijn rotsen in de branding: David (en Aukje), Martijn (en Altia), Reza (en

Caroline), Marloes, Ben, Strelitzia en Babs voor de tweede keer. Jullie hebben allen gemeen

wat ik het belangrijkst vind in het leven: lekker eten en lekker lachen met op zijn tijd een

serieus gesprek. Dank voor jullie fantastische vriendschap, jullie maken mij tot de persoon

die ik ben.

Mijn paranimfen Martijn en Babs, dank dat jullie op deze belangrijke dag in deze rol naast

mij staan.

De soon-to-be New York crew Marijana, Fatima, Babs (alweer?!); dank voor de altijd heerlijk

luchtige humor en de gezelligheid. Met wie anders rijd ik via de stoep om slagbomen heen

of breng ik een hele dag in een Starbucks door met zelf meegenomen eten.

Mijn overige collega-onderzoekers op de afdeling Verloskunde en Gynaecologie: Wendy,

Yvonne, Kim, Babette, Evelyne, Sylvia, Nienke, Emilie, Nicole, John, Sam, Marieke, Melek,

Averil, Nina, Manon en iedereen die ik vergeten ben. Bedankt voor de zeer geslaagde

congressen (Miami 2011!), de woensdagochtendkoffies, de borrels (ook al was ik er niet

vaak) en natuurlijk die fantastische ski-trip!

Als laatste wil ik mijn familie bedanken, mijn broer Jashvier, mijn zusje Renu en mijn moeder

voor hun niet aflatende steun.

227

228

Het drukken van dit proefschrift is mede mogelijk gemaakt door sponsoring van de vol-

gende partijen:

• de afdeling Verloskunde en Gynaecologie, Erasmus MC

• de Gemeente Rotterdam en het ‘Klaar voor een Kind’ programma

(www.klaarvooreenkind.nl)

• J.E. Jurriaanse Stichting

• Medical Dynamics

• Mpluz

• Danone

• BMA-Mosos

• Goodlife Pharma

• AbbVie BV

• Nutricia

• Vasa Previa Foundation (www.vasaprevia.nl)

“De Nederlandse Vasa Previa Foundation is een non-profit organisatie met als hoofddoel het

terugbrengen van babysterfte of blijvend letsel als gevolg van vasa previa. Enkele activiteiten

om dit te bereiken zijn het onder de aandacht brengen van het gevaar van vasa previa binnen

de relevante beroepsgroepen en het aansturen op aanpassing van de medische protocollen, die

belangrijk zijn voor diagnose en management van vasa previa. Daarnaast biedt de stichting

een luisterend oor en advies voor iedereen die vasa previa meemaakt of heeft meegemaakt.”

Virender Jashvant John Poeran was born at a gestational age of 38 weeks on March 21st, 1983 in the Diakonessen hospital in Paramaribo, Suriname, at exactly 6:40

AM. He came into this world in vertex presentation with a birthweight of 3560 grams, a height of 49 centimetres and a head circumference of 36.5 centimetres. The attending midwife granted him an Apgar score of 10/10.

At the age of 8 he moved to The Netherlands where he finished middle and high school. In the summer of 2001 he was accepted for medical school at Erasmus University, Rotterdam. During his medical training he stumbled upon the path of research in which he was guided by professor Steegers from the department of Obstetrics and dr. van der Meijden from the department of Dermatology. After obtaining his medical degree in 2008 he worked as a physician at the department of Obstetrics and Gynaecology at the Albert Schweitzer hospital in Dordrecht.

In 2009 he was accepted for a PhD-program in Perinatal Epidemiology under guidance of professors Bonsel and Steegers at Erasmus MC. He was additionally granted the possibility to obtain a Master’s degree in Epidemiology, which he finished in August 2012, the same year in which he decided to pursue a career in Psychiatry instead of Obstetrics and Gynaecology. At the end of his PhD-research he therefore worked as an emergency physician in Psychiatry at the Parnassia Bavo Group. However, in the fall of 2012 he fell in love with the city of New York and is currently working as a postdoctoral associate of Epidemiology at Weill Cornell Medical College in New York.